Meta Q4 Earnings: A New Era Driven by AI Agents

Posted on February 4, 2026June 30, 2026 by io-fund

Meta is a stock that our firm is watching this year with great anticipation. The headline results are good and important to review, yet what’s most important is what is up ahead as Meta begins to merge LLMs directly into their recommendation systems, a shift away from back-end optimizations based on algorithms. This is how management described the shift: “Our feeds will become more interactive overall. Today, our apps feel like algorithms that recommend content. Soon, you'll open our apps, and you'll have an AI that understands you and also happens to be able to show you great content or even generate great personalized content for you.”

There are many advertising businesses using AI, yet there is something unique about how Meta is approaching the problem – and this is showing up in the results. Per our Q1 2026 Top 15 AI Stocks report: “Margins matter, cash matters, but what matters more is the 3X growth Meta has seen in its Advantage+ segment in less than a year, as the company had reported $20 billion about three quarters ago, with the recent update from last quarter at $60 billion. If this runaway growth continues, then Meta will easily be outpacing Search and Google Cloud combined on AI revenue.”

That “something unique” is Meta’s data. What does Search really know about you other than intent? Meanwhile, Meta offers rich contextual, behavioural data that expands over time. The dataset offers preferences, emotions, social context, network effects, etc, whereas Google would struggle to truly “know you.” This can make the LLMs more personal and effective especially as we enter the agent era. Notably, this is not an either/or discussion about Google Search, rather it’s an opportunity to discuss Meta’s strengths and why the company is currently number two in AI revenue despite spending a less capex than its counterparts.

Below, I discuss the earnings call discussions that were centered around the plans Meta has for 2026 and also the numbers that drove strong after hour performance.

Meta to Replace Legacy, Rule-Based Algorithms with AI Agents

Meta is an “eyeballs” company, and thus, an important lever to growth is increasing user engagement. In the most recent quarter, the company drove incremental engagement from ranking and product improvements. Primarily, the company optimized their systems to consider longer interaction histories to better identify a person’s interests. This led to the highest lift in feed views that the company has seen in two years: “The optimizations we made in Q4 drove a 7% lift in views of organic feed and video posts on Facebook, resulting in the largest quarterly revenue impact from Facebook product launches in the past two years.”

Moving forward, Meta’s goal this year is to scale their training data to offer more personalized recommendations. By moving away from algorithms driving the feeds to LLMs, Meta can make he systems more responsive to real-time interest.

This may seem like a subtle shift, but it’s actually not subtle at all – Meta is proposing a complete overhaul in how their systems surface content. Moving forward, LLMs will offer reasoning for a level of personalization not possible in the current approach, which is more pattern recognition based. Think of how Spotify works – it surfaces music you’ve already listened to. Facebook feeds are similar. However, moving forward, Meta can offer a personalized agent approach to where AI optimizes a feed to suggest content that does not require a direct signal.

Here is what was stated on the call:

“We're seeing in our early testing that personalized responses drive higher levels of engagement, and we expect to significantly advance the personalization of Meta AI this year. This dovetails with our investments in content understanding, which will enable our systems to develop a deeper understanding of each person's interests and preferences while also identifying the most relevant content across our platform to pull into responses.”

Although Meta uses AI in its recommendations, the current systems are based on pattern and behavior-driven algorithms. For 2026, Meta will offer content that goes beyond the bounds of what you’ve already searched for/engaged with AI agents that can more intelligently infer your interests.

The result will be more time spent on the platform and with higher engagement. Even incremental gains here will lead to more advertising dollars.

Ad Platform Driven by Superintelligence

The second area that Meta is making “big bets” is the increasing monetization efficiency. Last quarter alone, the company doubled the number of GPUs used to train their GEM model for ads ranking. Similar to what was stated above, part of the improvements is using longer sequences of user behavior to inform the feed plus which ads are placed and when: “This new sequence learning architecture is significantly more efficient than our prior architectures which should enable us to further scale up the data, complexity and compute we use in our future ranking models to deliver performance gains.”

Meta’s main approach to increasing the effectiveness of ad placements remains user targeting, but just smarter user targeting. This results in 4X better results than using AI to increase overall ad load: “In fact, in the second half of 2025, our initiatives on Facebook to redistribute ads across users and sessions delivered a nearly 4x larger revenue impact than Facebook ad load increases.”

As you’ll see below in the Financials section, these improvements are making a material difference with Q4 revenue growing 17% QoQ and with a forward guide that implies the highest YoY growth rate for Meta since Covid-fueled 2021.

Driving Down Costs:

There are two primary ways that Meta plans to drive down costs. The first is to leverage AI internally to reduce their workforce.

Here is what was stated about using AI internally to replace engineers: “Since the beginning of 2025, we've seen a 30% increase in output per engineer with the majority of that growth coming from the adoption of agenetic coding, which saw a big jump in Q4. We're seeing even stronger gains with power users of AI coding tools, whose output has increased 80% year-over-year. We expect this growth to accelerate through the next half.”

The second is to use a mix of custom chips, lower-cost AMD GPUs and Nvidia GPUs to achieve their goals: “Procuring sufficient infrastructure capacity is central to these initiatives, and we're working to meet our silicon needs by deploying a variety of chips that optimally support each of our different workloads. To that end, in Q4, we extended our Andromeda ads retrievable engine, so it can now run on NVIDIA, AMD and MTIA. This, along with model innovations, enabled us to nearly triple Andromeda's compute efficiency. In Q1, we will extend our MTIA program to support our core ranking and recommendation training workloads in addition to the inference workloads it currently runs.”

The paragraph above spells out headwinds to Nvidia to where even if the market continues to grow, Nvidia’s overall percentage of the market will erode. We covered this in the Q3 2025 Top 15 AI Stocks report under the subheading “AI is Diversifying”

Financials

By Royston Roche

Q1 Revenue guide suggests fastest growth since Sept 2021

Q4 revenue grew by 23.8% YoY and 16.9% QoQ to $59.9 billion, beating estimates by 2.4%. Although strong sequential growth in Q4 is seasonal and Meta posted a 19.2% QoQ increase in Q4 2024, the current sequential growth is being achieved on a substantially higher revenue base of $51.2 billion versus $40.6 billion in the prior-year period. The strong revenue growth was primarily driven by robust demand stemming from AI advancements in ad recommendations, monetization, and user engagement.

Management issued strong revenue guide of $53.5 billion to $56.5 billion, implying a 30% YoY growth and a sequential decline of (8.2%) at the midpoint. While the QoQ contraction reflects normal seasonality, the implied 30% YoY growth represents the fastest pace in the last 4.5 years.

The company’s 2025 revenue grew by 22.2% YoY to $200.97 billion. Looking ahead, revenue growth is expected to accelerate 2.8 percentage points to 25% YoY growth to $251.3 billion in 2026 and will moderate to 16.9% YoY to $293.8 billion in 2027.

Record Q4 Advertising Revenue of $58.1 billion

Meta is already seeing tailwinds from AI recommendation models driving higher ROI for advertisers following increased time spent across its family of apps.

Q4 advertising revenue grew by 24.3% YoY to a record $58.1 billion. Notably, absolute advertising revenue growth reached $11.3 billion in the quarter, surpassing the $10.2 billion increase recorded in Q3.

Key Metrics

Q4 ARPP reaches a record $16.56

Perhaps the most important metric for Meta’s ad monetization is Family ARPP (average revenue per person). It reached a record $16.56 in Q4 2025, highlighting that Meta’s AI-driven ad performance improvements and monetization efforts are bearing fruit. While the 16.2% YoY growth in Q4 reflects a deceleration from the 17.7% seen in Q3, such a trend is common on a higher base and less of a concern given Meta is guiding for a modest acceleration this fiscal year. Notably, Q4 ARPP outpaced the 15.6% growth recorded in the prior-year period.

In Q4, the total number of ad impressions grew by 18% YoY, accelerating from 14% growth in Q3 and up from 6% in the year ago quarter. Impression growth was broad-based across regions, driven primarily by higher engagement and user growth, with incremental support from ad load optimizations.

Pictured Above: Ad impressions saw outsized growth this past quarter due to the new sequence learning architecture discussed above. Source: Meta investor relations.

The average price per ad continues to rise and it grew by 6% YoY in Q4, benefiting from increased advertiser demand, largely driven by improved ad performance. However, the YoY growth has decelerated from 10% in Q3 and 9% in Q2 and this metric is to be watched despite ad impressions offsetting the deceleration.

Family of Apps daily active people (DAP) grew by 6.9% YoY to 3.58 billion in Q4, though the growth rate decelerated slightly from 7.6% growth in Q3.

Margins

Meta delivered a sequential improvement in operating margin in Q4 as its continued investments in AI are beginning to show early signs of payoff, even though margins were lower than a year ago quarter.

Q4 gross margin was 81.8%, up 10 basis points YoY and down 20 basis points sequentially.
Q4 operating income increased 5.9% YoY to $24.7 billion, with an operating margin of 41.3%. The margin improved by 130 basis points sequentially, but declined 700 basis points from a year ago, largely due to higher AI-related operating expenses.
Looking ahead to 2026, management expects total expenses to grow by 40.6% YoY to $165.5 billion—driven largely by increased infrastructure spending and continued investment in AI talent. Even with those higher costs, the company still expects operating income to grow in 2026. Management also emphasized that losses at Reality Labs are not expected to increase next year.
Q4 net income grew by 9.3% YoY to $22.77 billion with a net profit margin of 38% compared to 43.1% in the same period last year.

GAAP EPS beat of 8%

Q4 GAAP EPS grew by 10.7% YoY to $8.88, beating estimates by 8%, driven primarily by higher revenue from stronger AI monetization. Analysts expect EPS to grow 2.4% YoY to $6.59 in Q1 2026 and down (0.3%) YoY to $7.12 in Q2 2026.

Looking ahead, GAAP EPS is expected to grow 26.5% YoY to $29.71 in 2026 and 16.3% YoY to $34.55 in 2027.

Cash Flow and Balance Sheet

Meta’s cash flows improved in Q4 driven by higher profits.

Q4 operating cash flow grew by 29.4% YoY to $36.2 billion with an operating cash flow margin of 60.5% compared to 57.8% in the same period last year.
Q4 free cash flow grew by 7% YoY to $14.1 billion with a free cash flow margin of 23.5% compared to 27.2% in the same period last year. Q4 capex grew by 49.2% YoY to $22.14 billion. 2025 capex grew by 84.1% YoY to $72.22 billion. Management expects 2026 capex to be $115 billion to $135 billion, implying a YoY growth of 73.1% at the midpoint primarily due to higher AI investments.
The company had cash & marketable securities of $81.6 billion and debt of $58.7 billion compared to $44.45 billion and $28.8 billion at the end of Q3. The company issued debt of $30 billion in Oct 2025. Meta also entered a joint venture with Blue Owl Capital to fund its development at the Hyperion data center in Louisiana. Thereby, helping it to keep about $27 billion in debt off-balance sheet, where it would sit in a special-purpose vehicle tied to Blue Owl. While this approach may improve reported leverage and financial ratios, it carries inherent risks as the company is indirectly responsible for the off-balance sheet debt.

Conclusion

Meta was able to put up strong results primarily by increasing ad impressions, despite seeing a decelerating average price per ad from 14% growth last year to 6% growth this year. The combination resulted in average price per person reaching $16.56 up from $14.25 in the year ago quarter.

Facebook feeds saw an improvement in its sequencing learning architecture. In other words, Meta tracks user history longer to infer context on what the user is most interested in seeing next, which led to higher user engagement and a spike in ad impressions. Early signs of this improvement are evident with a forward guide that implies the highest YoY growth rate for Meta since 2021 at 30% YoY growth guided for Q1.

Most importantly, management is stating there will be a substantial shift to their underlying recommendation systems by re-architecting its traditional algorithms that are more pattern-based to now become more intuitive and forward-thinking in terms of anticipating what a user will want to see next.

Royston Roche, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the company’s portfolio. We then share that information with our readers and offer real-time trade notifications. This is not a guarantee of a stock’s performance and it is not financial advice. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis. Beth Kindig and the I/O Fund own shares in META at the time of writing and may own stocks pictured in the charts.

Recommended Reading:

Meta: Growth is Quietly Benefitting from AI, Though Margin Risks Weigh Heavy

Posted on December 17, 2025June 30, 2026 by io-fund

While much attention is given to Nvidia and the AI semiconductor ecosystem for visible AI-driven hypergrowth trajectories, there is another quiet AI beneficiary emerging in Meta, with advertising revenue accelerating to the mid-20% range and YoY dollar growth surpassing $10 billion in Q3. Revenue forecasts continue to strengthen over the next few years with revisions of up to $30 billion, or up 10-14% since July, underscoring greater confidence in Meta’s ability to leverage AI to improve monetization.

More impressively, Meta’s AI ads automation platform has quietly reached a $60 billion run rate in Q3 in three and a half years from its launch. This is 3X Broadcom’s AI revenue and also 3X of OpenAI’s projected ARR of $20 billion for 2025, emphasizing how large of a platform Advantage+ is. Compared to Microsoft, AI contributed 16 points of growth in the April quarter, implying a run rate of around $16 billion assuming growth at a similar degree as the January quarter’s 175% YoY to $13 billion.

However, the thorn in Meta’s side stems from the compute and capex side, as the company is aggressively building data center capacity to prepare itself for the most optimistic scenarios of reaching superintelligence. Not only is capex guided to surge to over $100 billion in 2026, potentially creating another cash flow crunch reminiscent of 2022’s metaverse-linked spending spree, but expense growth is also expected to outpace revenue growth by a wide degree and weigh heavily on operating margin.

Advantage+ Reaches $60 Billion ARR

Meta’s flagship AI ads automation platform Advantage+, powered by Andromeda, is also key to the company’s strong ads performance, as Meta has been straightforward about the end-to-end platform driving strong return on ad spend for advertisers. Advantage+ automates campaign targeting, budget allocation, and creative generation, providing advertisers with an easy-to-use tool that integrates generative AI directly into Meta’s ad ecosystem.

Meta revealed earlier this year that “for every dollar spent on its AI-enabled Advantage+ products, advertisers generate on average $4.52 in revenue for their businesses,” or an increase of ~22% versus typical campaigns, highlighting how ad performance improves while using the platform.

In Q3, Meta emphasized that “Advantage+ continues to drive performance gains, [and] advertisers who run lead campaigns using Advantage+ are seeing a 14% lower cost per lead on average than those who are not.” This compares to a 10% lower cost per lead as of April, showing that the platform continues to drive results while lowering costs per lead.

More importantly, Meta disclosed that Advantage+ has surpassed a $60 billion annual run rate, just three and a half years after its launch. CFO Susan Li sees room to continue growing this run rate for Advantage+ and expanding adoption of the platform by focusing on driving continued performance improvements.

First Monetization Lever: Improving Ad Performance

Meta outlined two monetization levers in Q3 – improving ad performance, and an ability to continue delivering engaging content to users.

This first monetization lever stems from improving ad performance for its advertisers, mostly driven by the company’s three foundational models as well as its end-to-end ads automation platform Advantage+. Meta opts for tracking conversions to gauge ad performance, despite it being a complex metric to track considering advertisers can optimize for different types of conversions. CFO Susan Li stated that value-weighted conversion rates showed “very strong” YoY growth in Q3, outpacing impressions.

Meta’s three foundational models all serve a different function, with the same end goal of improving ad quality and conversions to drive higher ROI for advertisers:

GEM (Generative Ads Recommendation Model) is described by Meta as the ‘super brain’ that can rapidly process, catalog and analyze trillions of data points, to then recognize subtle patterns in user activity to provide the most relevant ads at the right time. Meta says GEM was rolled out more broadly earlier this year after initial testing on Reels saw GEM boost conversions by up to ~5%. GEM delivered a 5% increase in conversions on Instagram and a 3% increase on Facebook in Q2, and in Q3 Meta “doubled the performance benefit we get from adding a given amount of data and compute” to continue scaling training capacity at an attractive ROI.
Lattice is described as a ‘giant library’ that generalizes learnings across different campaign objectives (clicks, views, etc), surfaces (Reels, Story, Feed, etc) and subjects, in order to predict an ad’s performance. Lattice increases ad efficiency as it runs fewer models, while the knowledge-sharing effect increases ad quality and conversions – Meta said earlier this year that Lattice has increased ad quality by 12% and conversions by 6%. In Q3, Meta rolled out Lattice to app ads, driving a ~3% gain in conversions on that objective.
Andromeda is described as a ‘personal concierge’, or Meta’s vast ML ad recommendation and prediction system that, at its core, aims to predict exactly which ads a user will find the most interesting. For Andromeda, Meta says, “Imagine having a personal concierge who knows your tastes so well that they don’t just understand that you covet shoes, but that you like to wear red flip flops at the beach.” Meta said that in Q3, it significantly improved Andromeda’s performance by combining retrieval and early-stage ranking models, driving a 14% increase in ad quality on Facebook. Andromeda is also the core engine powering Advantage+ automation tools.

Moving to 2026, Meta discussed that it is “working on combining these 3 major AI systems into a single unified AI system that will effectively run our family of apps and business using increasing intelligence to improve the trillions of recommendations that it will make for people every day.”

A single model that combines the strengths of GEM, Andromeda and Lattice could theoretically understand user preferences and activity at a much deeper level, improve ad ranking quality, relevance and conversions across its family of apps, and save on inference. For example, Meta does not use GEM for inference as its size makes it too cost-prohibitive, rather it transfers knowledge to smaller run-time models; a single model incorporating GEM’s knowledge could potentially run inference in a more cost-effective manner.

Second Monetization Lever: Increasing User Engagement

On the second of increasing engagement, Meta is executing quite well, with improvements in its recommendation models helping drive time spent on its apps higher. More time spent then allows ad impressions to grow without substantially increasing ad load, underpinning this reacceleration in impressions growth seen in Q3 and more growth moving forward.

Management pointed out that “overall time spent on Facebook and Instagram grew double digits year-over-year, driven by continued video strength as well as healthy growth in nonvideo time on Facebook.” Video time spent on Instagram was more than 30% higher versus last year, while AI ranking optimizations helped drive 10% more time spent on Threads in Q3. This video growth has pushed Reels to a $50 billion annual run rate in Q3, up 5X from its last update in Q2 2023 when it reached a $10 billion run rate.

Improving ranking models remains a key focus for Meta moving through 2026, with management expecting new model innovations to help “significantly scale up the amount of data and compute we use to train our recommendation models in 2026, yielding more relevant recommendations.”

AI Aiding Meta’s Advertising Growth Flywheel

On a positive note, Meta is already seeing tailwinds from AI recommendation models driving higher ROI for advertisers and increasing time spent across its family of apps, fueling stronger advertising revenue growth.

In Q3, advertising revenue grew 25.6% YoY, accelerating more than nine points since Q1 and marking the fastest growth in six quarters. Ad impressions rose 14% YoY in Q3, accelerating from 11% in Q2 and marking a strong inflection from just 5% growth in Q1. Pricing remained steady, rising just one point to 10% YoY.

However, the dollar growth in advertising stands out more — Meta has delivered its two largest YoY growth quarters on a dollar basis in Q2 and Q3, at $8.23 billion and $10.2 billion, even outperforming Q4 2024’s holiday-boosted growth of $8.08 billion. This high dollar growth is poised to continue in Q4 2025, with guidance pointing to ~$9.3 billion to $10.7 billion in QoQ dollar growth.

Put another way, Meta is delivering larger YoY dollar growth in advertising revenue on a larger base – Q3 grew $10 billion YoY off a $40 billion base, versus $7.5 billion growth in Q3 2024 on a $33.6 billion base.

ARPP Continues to Accelerate Heading into Q4

Perhaps the most important metric for Meta’s ads monetization is ARPP (average revenue per person), with the metric continuing to accelerate in Q3 ahead of the seasonally-stronger holiday quarter. ARPP reached $14.46 in Q3, accelerating to 17.7% YoY from 14.8% in Q2. More impressively, this marked a record high for ARPP, surpassing Q4 2024’s seasonally stronger ARPP of $14.25.

This sets the stage for ARPP to push well beyond $15, potentially to $16 in the upcoming quarter, highlighting that Meta’s AI-driven ad performance improvements and monetization efforts are bearing fruit.

Meta’s Upcoming Capex Surge and Possible FCF Crunch

Some of the most important quotes from Q3’s call circled back to Meta’s view on capex and why it believes aggressive expansion of capacity and thus capex is a necessity. CEO Mark Zuckerberg believes it is the “right strategy to aggressively front-load building capacity so that way we're prepared for the most optimistic cases” on when AI superintelligence arrives, so Meta is prepared to capitalize on this opportunity.

If building superintelligence takes years longer than expected, Zuckerberg says Meta can “use the extra compute to accelerate our core business which continues to be able to profitably use much more compute than we've been able to throw at it. And we're seeing very high demand for additional compute, both internally and externally.

These comments underscore why Meta is aggressively raising its capex spending this year and next – Zuckerberg believes that the upside potential of superintelligence is so large that it is worth the risk of overbuilding to not fall behind OpenAI or Google (with compute capacity being the main advantage), with Meta able to use extra compute in the meantime to improve core AI ad capabilities and drive growth.

However, the tradeoff for this is lower free cash flow and potential operating margin headwinds. Meta expects capex dollar growth to be “notably larger in 2026 than 2025,” while total expenses “will grow at a significantly faster percentage rate in 2026” driven by infrastructure costs, incremental cloud costs and depreciation, followed by employee compensation.

This implies 2026 capex of at least $103 billion, as current guidance for 2025 at $70-72 billion implies a minimum of ~$32 billion YoY growth. However, considering management’s comments for notably larger dollar growth, there is potential for capex to come in at or above $110 billion, up ~55% YoY, above current estimates for $107.9 billion. Put another way, Meta could spend ~$30 billion more in 2025 and 2026 than it did in 2019 through 2024 combined.

The capex surge will potentially cause another free cash flow crunch similar to 2022, with current consensus estimates pointing to FCF of $19.71 billion in 2026, down nearly (50%) YoY and (63.5%) from 2024.

Expense Growth to Meaningfully Outpace Revenue Growth in 2026

Tying into capex is Meta’s expectation for expense growth to be significantly faster in 2026 versus 2025, which means expense growth will outpace revenue growth by a wide margin, potentially as much as a factor of 2x.

For perspective, Meta is forecasting total operating expenses of $116-118 billion this year, up 22-24% YoY, marginally outpacing estimated revenue growth of 21.3%. This is crucial heading into next year as a lack of meaningful gross margin expansion means this growth will directly pressure operating margin.

If expenses grow at ~30% YoY in 2026, this would project total expenses in the range of ~$151-153 billion, significantly outpacing expected revenue growth of 17.9% to $234.7 billion. This would also project out to an operating margin of ~35%, marking a relatively sharp contraction back to the lowest levels since early 2023.

If expenses grow faster, at say 2X estimated revenue growth or ~35% YoY, this would project expenses in the range of ~$157-159 billion. While only slightly higher than the ~30% growth forecast, this would bring operating margin down to 32.5%-33%.

JP Morgan’s Doug Anmuth questioned management about this capacity expansion strategy and how this spending ties to earnings and cash flow:

“I appreciate the strategy to front load capacity for superintelligence. Can you just talk about your thought process and kind of triangulating the Capex dollar growth and the significantly faster expense growth next year with core growth in the business and then the impact on earnings and free cash flow? And do you have targets that we should be thinking about for cash on hand or net cash overall?”

CFO Susan Li offered a lengthy discussion in response that offered no clarification on earnings or cash flow impacts, and hinted that Meta may not be worried about those two line items in this buildout:

“But to date, we keep on seeing this pattern where we build some amount of infrastructure to what we think is an aggressive assumption. And then we keep on having more demand to be able to use more compute, especially in the core business in ways that we think would be quite profitable than we end up having compute for.

So I think that, that suggests that being able to make a significantly larger investment here is very likely to be a profitable thing over some period… Now I mean, it's, of course, possible to overshoot that, right? … And then the kind of the very worst case would be that we effectively have just prebuilt for a couple of years, in which case, of course, there would be some loss and depreciation, but we'd grow into that and use it over time.”

On the point of Meta’s 2026 budget still being put into place, reports surfaced that Meta is considering making budget cuts of up to 30% in its metaverse division, Reality Labs, which is currently burning about ~$20 billion per year. This could save several billion if put into place, though there is potential for that money to simply be reallocated towards data center capex.

Financials

Revenue Growth Accelerates to 26%

Meta reported revenue of $51.24 billion in Q3, accelerating more than 4.5 points to 26.2% YoY, the highest growth since Q1 2024. This also marked an impressive reacceleration from 16% growth in Q1 2025.

For Q4, management guided to revenue between $56 to $59 billion, up 18.8% YoY at midpoint, driven by expectations for strong ad revenue growth, partially offset by lower YoY revenue in Reality Labs from lapping the Quest 3S introduction. For 2025, revenue is expected to grow 21.3% to $199.5 billion, before decelerating to 17.7% growth to $234.7 billion in 2026.

Annual Revenue Revisions Seeing Sharp Increase Since July

What’s notable on the revenue front is the sharp upward revisions to annual revenue estimates, with 2026 and 2027 moving sharply higher since this summer.

Back in July, prior to Q2’s earnings, Meta was expected to generate $215.1 billion in revenue, with that now sitting at $234.7 billion. On a YoY basis, growth has been revised from 14.0% to 17.7%, a smaller uplift considering 2025 comps have toughened, having risen from 14.7% to 21.3% over the same period.

For 2027, Meta was expected to generate $240.6 billion in late July, with that now sitting at $271.0 billion, with YoY growth moving from 11.9% to 15.5% on a higher base.

Key Metrics

Meta’s key metrics strengthened broadly in Q3. Ad impressions growth accelerated three points to 14% YoY, its fastest growth since Q1 2024. Pricing has remained relatively stable at 10% YoY in Q3, up one point sequentially, driven by increased advertiser demand fueled by improved ad performance.

Family of apps daily active people (DAP) also accelerated to 7.6% YoY to 3.54 billion, up from 6.4% growth last quarter and marking its fastest growth since Q4 2023.

Operating Margins Contracts Sequentially

Despite the topline reacceleration in Q3, Meta’s operating margin contracted as expenses grew 32% YoY, six points faster than revenue growth.

Gross margin was 82% in Q3, up 0.2 points YoY but down 0.1 points QoQ.
Operating margin was 40%, down 2.7 points YoY and 3.0 points QoQ. Aside from Q4 typically being seasonally stronger, the expense guide for next year suggests operating margin could return to the mid-30% range.
Net margin was 5.3%, negatively impacted by a one-time, non-cash income tax charge of $15.93 billion; excluding this charge, net margin would’ve been 36.4%, down 2.3 points YoY and 2.2 points QoQ.

Earnings

Due to the income tax charge, Meta reported $1.05 in GAAP EPS; adjusted for this charge, EPS was $7.25, compared to estimates for $6.67.

Looking ahead, GAAP EPS growth is expected to remain approximately flat for the next three quarters due to the margin pinch from rising expenses:

Q4 GAAP EPS estimated at $8.17, up 1.9% YoY.
Q1 ’26 GAAP EPS estimated at $6.32, down (1.7%) YoY.
Q2 ’26 GAAP EPS estimated at $7.08, down (0.8%) YoY.

For FY25, Meta is expected to deliver a (2.2%) decline in GAAP EPS to $23.34, with the decline primarily due to Q3’s income tax charge-related miss. Earnings growth is expected to rebound to 27.3% to $29.72 in 2026.

Cash and Balance Sheet

Operating and free cash flow margins expanded sequentially, though Meta is expected to see a steep free cash flow crunch moving through 2026 as a result of surging capex (which also does not reflect the company’s true spending on data center infrastructure).

Operating cash flow was $30.0 billion in Q3 for a 58.5% margin, down from a 60.9% margin in the year ago quarter but up from a 53.8% margin in Q2.
Free cash flow was $10.63 billion for a 20.7% margin, down sharply from a 38.2% margin in the year ago quarter but up from 18% in Q2.
Q3 capex rose 110.5% YoY to $19.34 billion, driven by investments in servers, network infrastructure and data centers. Based on 2025’s capex guide, which was raised to $70-72 billion (up 81% YoY at midpoint), Q4 capex is on track to be ~$21 billion, up 41.5% YoY.

Creative Funding Solutions Not Appearing in Capex, Saving Cash

Another important point to cover is Meta’s use of creative financing solutions to build out its large scale data centers, such as its joint venture with Blue Owl to fund its Hyperion data center in Louisiana. The additional costs are not appearing in capex but rather in ‘other investing cash flows’.

Under the JV deal, Blue Owl will own 80% and Meta will retain a 20% stake, overseeing construction and ultimately renting the data center once operational. However, concerns are rising about the deal structure, as the WSJ points out that the facility is “financed with debt, and neither the data center nor the debt will be on [Meta’s] own balance sheet.”

Truist analysts questioned CFO Susan Li about the JV and how it will appear in and affect capex:

“And then Susan, how do you see the on-balance sheet versus off-balance sheet financing of your AI initiatives? You've recently struck a deal with Blue Owl for the Louisiana data center. Is that part of the CapEx guide for '26? And if it's not, how significant will that way of funding be for Meta going forward? And basically, would that slow down your CapEx growth past 2026?”

CFO Susan Li: “So the JV that we announced with Blue Owl is sort of an example of finding a solution that enabled us to partner with external capital providers to codevelop data centers in a way that gives us long-term optionality in supporting our future capacity needs just given both the magnitude, but also uncertainty of what the capacity outlook in future years looks like.

In terms of how that is recognized as Capex, our prior Capex reflected a portion of the data center build cost prior to the joint venture being established. Going forward, the construction cost of the data center will not be recorded in Capex as the data center is constructed, we will contribute 20% of the remaining construction costs required, which is in line with our ownership stake, and those will be recorded as other investing cash flows.”

Valuation

Meta’s shares are trading slightly above its median forward PS valuation since the start of 2024 at 8.1x, though this has compressed from 9.4x in prior to Q3 earnings in late October after shares sold off.

On the bottom line, Meta is valued at around a 25x forward PE, slightly above its 5-year median of 22.2x, though the company had traded as low as the single-digits in late 2022 and early 2023 when its metaverse spending spree cut into operating and net margins. Looking out to 2026, Meta trades at ~22x current EPS estimates, slightly above its average of 20.7x since the start of 2024.

However, where the valuation gets stretched is on the free cash flow side – looking ahead to 2026 and the projected $19.7 billion in free cash flow (subject to change with capex forecasts), Meta trades at 85x estimated 2026 FCF. This would represent a significant deterioration of this multiple from the current 38.3x and represent the most expensive Meta has traded on an FCF basis since shortly after its IPO.

Conclusion

The most recent earnings report proves that Meta is using AI internally to materially move the needle. Meta’s Advantage+ automation tools continue to drive measurable improvement in advertising efficiency with an updated annual run rate that exceeds $60 billion – a substantial run rate considering this was launched only three years ago.

Across the board, there was a noticeable reacceleration in impressions, and ARPP was at a fresh record despite it not being the company’s seasonally strongest quarter. Plus, Reels is also up 5X in two years, reporting a $50 billion run rate following improvements in the AI-driven video content recommendation system.

However, the main risks to Meta’s thesis lay within its ambitious capacity expansion plans, with management laying the framework for easily over $100 billion in capex in 2026 and notably stronger expense growth. This is not only expected to create a large FCF crunch similar to 2022, back to <$20 billion, but also a substantial headwind to operating margins and thus EPS.

There’s an ongoing glass-half-full versus glass-half-empty debate in AI — enormous potential on one side, significant cost on the other. We remain firmly in the glass-half-full camp. But if the market chooses a glass-half-empty view, we also know that lower prices often create some of the best long-term opportunities.

Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.

Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.

Recommended Reading:

Alphabet Q1 2024: Impeccable Earnings Report, GCP Accelerates from AI

Posted on May 10, 2024June 30, 2026 by io-fund

Alphabet’s revenue accelerated for the sixth consecutive quarter. Revenue grew by 15% and clocked the fastest growth since Q1 2022. Google Cloud revenue accelerated for the second consecutive quarter. The company also announced its first dividend and authorized a new $70 billion share repurchase plan. The margin improvement due to the cost reduction initiatives was the icing on the cake. The shares closed 10% higher following strong results and reached the $2 trillion market capitalization milestone.

Revenue

Revenue grew by 15% and 16% in constant currency YoY to $80.54 billion and beat estimates by 2.3%. Revenue accelerated for the sixth consecutive quarter. Analysts expect revenue to grow 12.5% YoY to $83.90 billion in the June quarter. While growth is decelerating, the consensus estimates have moved up 100 basis points compared to the estimates prior to the earnings.

Google Services segment revenue grew by 14% YoY to $70.4 billion. Search and other advertising revenues accelerated to 14% YoY growth to $46.2 billion from 13% in the December quarter and 11% in the September quarter, which was primarily due to strong growth in retail, particularly from APAC-based retailers. The strong trend in the advertising business from APAC-based retailers began in Q2 2023 last year and continued in the recent quarter.

Management also highlighted tougher comps in the upcoming quarter and headwinds from the strong dollar. “As we look ahead, two points that will affect sequential year-on-year revenue growth comparisons across Alphabet. First, Q1 results reflect the benefit of leap year, which contributed slightly more than one point to our revenue growth rate at the consolidated level in the first quarter. Second, at current spot rates, we expect a larger headwind from foreign exchange in Q2 versus Q1.”

Margins

The company’s cost control initiatives, such as workforce reductions and office space optimizations, have helped the company report very strong margins in the recent quarter. Going forward, management expects operating margin for 2024 to be higher than in 2023. This is from moderating expense growth for higher depreciation and expenses related to higher technical infrastructure investments.

Gross margin expanded 200 basis points YoY and 160 basis points QoQ to 58.1%.
Operating margin expanded 660 basis points YoY and 410 basis points sequentially to 31.6%. Non-GAAP operating margin, which excludes severance and related office space charges, came in at 32.5% compared to 28.6% in the same period last year.
The operating margin was the second highest in the previous 10 years, and if we exclude the non-recurring charges in the recent quarter, it is the highest in the last decade.

Ruth Porat, CFO of the company said in the earnings call:

“Turning to margins, our efforts to durably re-engineer our cost base are reflected in a 400 basis point expansion of our Alphabet operating margin year-on-year, excluding the impact of restructuring and severance charges in both periods. You can also see the impact in the quarter-on-quarter decline in headcount in Q1, which reflects both actions we have taken over the past few months and a much slower pace of hiring. As we have discussed previously, we are continuing to invest in top engineering and technical talent, particularly in Cloud, Google DeepMind, and technical infrastructure. Looking ahead, we remain focused on our efforts to moderate the pace of expense growth in order to create capacity for the increases in depreciation and expenses associated with the higher levels of investment in our technical infrastructure. We believe these efforts will enable us to deliver full-year 2024 Alphabet operating margin expansion relative to 2023.”

–End Quote

Net margin improved 780 basis points YoY and 540 basis points sequentially to 29.4%. GAAP EPS came in at $1.89, up 61.5% YoY and beat estimates by 25%. It partly benefitted from a net gain on equity securities of $2.2 billion, and the $104 million reversal of previously accrued performance fees related to some of these investments, which increased net income by $1.9 billion and EPS by $0.15.

Analysts expect the company to report GAAP EPS of $1.83 in the next quarter, up from expectations of $1.69 prior to the earnings report.

Cash Flows and Balance Sheet

Operating cash flow was $28.85 billion or 35.8% of revenue compared to $23.51 billion or 33.7% in the same period last year. Free cash flow was $16.84 billion or 20.9% of revenue compared to $17.22 billion or 24.7% in the same period last year. Free cash flow was lower than the previous year as capex increased 91% YoY to $12 billion due to AI.

The CFO said in the earnings call, “With respect to CapEx, our reported CapEx in the first quarter was $12 billion, once again driven overwhelmingly by investment in our technical infrastructure with the largest component for servers followed by data centers. The significant year-on-year growth in CapEx in recent quarters reflects our confidence in the opportunities offered by AI across our business. Looking ahead, we expect quarterly CapEx throughout the year to be roughly at or above the Q1 level, keeping in mind that the timing of cash payments can cause variability in quarterly reported CapEx.”

This means that capex is expected to rise about 49% YoY to $48 billion in 2024.

Cash and marketable securities were $108.09 billion compared to $110.92 billion in the December quarter. Debt was also largely unchanged at $13.23 billion compared to $13.25 billion in the December quarter.

Another key highlight in the report was the announcement of the first quarterly dividend of $0.20 and the authorization of an additional $70 billion share repurchase plan. The company repurchased shares worth $15.7 billion in the recent quarter.

Key Metrics:

Google Cloud Revenue

Google Cloud revenue grew by 28% YoY to $9.6 billion, helped by increasing contributions from AI and strong Workspace growth, which is productivity apps. The revenue growth accelerated for the second consecutive quarter from 26% in the previous quarter and 22% in the September quarter. The strong growth also further helped to narrow the gap with Microsoft Azure’s growth of 31%. Google Cloud now only trails Microsoft Azure by 3 percentage points compared to 4 points and 7 points in the previous two quarters.

The operating margin for Google Cloud came in at 9% compared to 3% in the same period last year and 9% in the December quarter.

Google Advertising Revenue

Google Advertising revenue accelerated for the fifth consecutive quarter. Revenue grew by 13% YoY to $61.7 billion, compared to 11% growth in the December quarter, and was flat in the same period last year.

Google Search and other advertising revenues grew by 14% YoY to $46.2 billion. It was up from 13% growth in the previous quarter and 2% in the same period last year.
YouTube ads revenue grew by 21% YoY to $8.09 billion, helped primarily by direct response marketing and brand advertising. It was up from 16% growth in the previous quarter and a decline of (-3%) in the same period last year.
Networking advertising revenue declined by (-1%) YoY to $7.41 billion. It was better than the (-2%) decline in the previous quarter and (-8%) decline in the same period last year.

Earnings Call:

Capex

Big Tech capex has surged over the past few quarters and Google is no exception.

Capex grew 91% YoY to $12 billion; this is up from 45% growth and $11 billion last quarter. As stated, this means that capex is expected to rise about 49% YoY to $48 billion in 2024.

The CFO stated less than 10% was going to office with the rest going toward infrastructure (or AI basically).

Per the earnings call:

Ruth Porat:

“And then in terms of CapEx, as I said in opening comments, we do expect the quarterly CapEx throughout the year to be roughly at or above the $12 billion cash CapEx we had here in Q1. As I said, you can always have variability in the reported quarterly CapEx just due to the timing of cash payments, but roughly at or above this level. And it really goes to Sundar's comment, opening comment, that we're very committed to making the investments required to keep us at the leading edge in technical infrastructure to support the growth in Cloud, all the innovation and search that he and Philip have spoken about and our lead with Gemini. I will note that most nearly all, I should say, of the CapEx was in our technical infrastructure. We expect that our investment in office facilities will be about less than 10% of the total CapEx in 2024, roughly flat with our CapEx in 2023, but is still there […]

Clear Path to Monetization

In the call, the CEO highlighted six points as to why the company is positioned to tap the AI opportunity.

Leadership position in R&D
Infrastructure leadership including Google Cloud
The company invests in developing new AI models; custom TPUs power AI projects which are in the 5th generation and this powers Search
Global product footprint
Velocity in execution (my note: this one can lag at times)
Monetization paths

Of these, the last one on monetization paths interests us the most.

Regarding #4, the global product footprint combined with infrastructure leadership (#2) will help with data sovereignty in the medium-term as AI will become required to have data residency in the country the data is produced. This isn’t unique necessarily as all hyperscalers offer this, but it’s notable and supports our thesis that Big Tech will get Bigger.

The clear path to monetization that Google speaks about is through Google Cloud, which accelerated this quarter and is closing the gap with juggernaut Azure. Secondly, it’s through Search revenue and ad revenue on YouTube and its advertising network. As stated above, Google Cloud accelerated to 28% growth, up from 26% growth in the previous quarter. Search also accelerated to 14% growth this quarter for revenue of $46.2 billion, compared to growth of 13% and revenue of $48 billion last quarter (seasonal high due to being Q4). These are record quarters for Search revenue.

Per the opening remarks – note, SGE refers to Google Search Generative Experience (SGE) which uses gen AI to provide brief overviews of web pages.

“We have clear paths to AI monetization through ads and cloud, as well as subscriptions. Philip will talk more about new AI features that are helping advertisers, including bringing Gemini models into Performance Max. Our Cloud business continues to grow as we bring the best of Google AI to enterprise customers and organizations around the world. And Google One now has crossed 100 million paid subscribers. And in Q1, we introduced a new AI premium plan with Gemini Advanced […]

“You can see that from the increases in our capital expenditures. This will fuel growth in cloud, help us push the frontiers of AI models, and enable innovation across our services, especially in Search. AI innovations and Search are the third and perhaps the most important point I want to make. We have been through technology shifts before, to the web, to mobile, and even to voice technology. Each shift expanded what people can do with Search and led to new growth. We are seeing a similar shift happening now with generative AI. For nearly a year, we have been experimenting with SGE and Search labs across a wide range of queries. And now we are starting to bring AI overviews to the main Search results page. We are being measured in how we do this, focusing on areas where gen AI can improve the Search experience, while also prioritizing traffic to websites and merchants.

We have already served billions of queries with our generative AI features. It's enabling people to access new information, to ask questions in new ways, and to ask more complex questions. Most notably, based on our testing, we are encouraged that we are seeing an increase in Search usage among people who use the new AI overviews as well as increased user satisfaction with the results.”

–End Quote

Later, the CFO added: “We're continuing to experiment with new ad formats, including search and shopping ads alongside search results in SGE. And we shared in March how folks are finding ads either above or below the SGE results helpful. We're excited to have a solid baseline to keep innovating on and confident in the role SGE, including ads, will play in delighting users and expanding opportunities to meet user needs.”

Toward the end of the call, an analyst asked the CEO to quantify the monetization. He did not provide specifics, rather stated: “There are questions about monetization, and based on our testing so far, I am comfortable and confident that we'll be able to manage the monetization transition here well as well.”

My note: there is a lot of corporate-speak on these calls to wade through. However, to keep it brief and direct, areas where Google can continue to monetize Search is on mobile with rumours that Apple may partner with Google’s Gemini to power AI apps on iPhones in the future. Android also currently has 3 billion users that Gemini can reach.

60% of Funded Gen AI Startups use Google Cloud

Google Cloud has the majority of funded AI startups building on Google Cloud. Although this is too small to make a revenue impact, the message is that Google Cloud is at the cutting edge for AI workloads training on Nvidia’s GPUs and its own TPUs. Google also recently announced Axion CPUs.

“Today, more than 60% of funded Gen AI startups and nearly 90% of Gen AI unicorns are Google Cloud customers […] We also announced Axion, our new Google design and ARM-based CPU. In benchmark testing, it has performed up to 50% better than compatible x86-based systems. On top of our infrastructure, we offer more than 130 models, including our own models, open source models, and third-party models. We made Gemini 1.5 Pro available to customers, as well as Imagine 2.0 at Cloud Next.”

It was also stated that over 1 million developers use Google’s generative AI tools.

Google Third-Party Cookie Deprecation

Google has delayed the third-party cookie deprecation on its browser for the third time with plans to now phase them out in 2025. As of now, it has restricted third-party cookies to only 1% of Chrome users. The delay is due to concerns from the regulatory authorities, particularly the UK’s Competition and Markets Authority (CMA) and the ad industry. The recent release said, “It's also critical that the CMA has sufficient time to review all evidence including results from industry tests, which the CMA has asked market participants to provide by the end of June.”

Regulators like the UK’s Competition and Markets Authority (CMA) have previously raised concerns that Google’s plan to replace third-party cookies with its Privacy Sandbox initiative will give Google an unfair advantage in the ad market. According to eMarketer and 33Across, cookies were used in 78% or more of programmatic ad buys. Google’s Privacy Sandbox replaces cookies with 20 different APIs for “Measurement and Relevance” and also “Other APIs” to help advertisers target users based on topics and cohorts without tracking individual behavior. For users and advertisers, this is an improvement from cookies. For competitors like The Trade Desk, this poses a threat as Google own major properties across the web, and theoretically, eliminating cookies can limit the effectiveness of a platform like The Trade Desk. This is because cookies leak a lot of data to third parties like The Trade Desk, whereas what Google seeks to do is stop those leaks to outside parties on their own properties (Chrome for now, Android soon after).

Despite Google being the target for alleged monopolistic control over Search and its ad network, Apple and Mozilla have already phased out identifiers on their browsers, and Apple has also done this on iOS. We covered this extensively, including here. According to Google, Privacy Sandbox will use the same Android API solutions, and no one (including Google) will have a privileged position (given the recent allegations in the DOJ lawsuit, I’d take that with a grain of salt).

Ad-tech companies like The Trade Desk have been critical of Google’s Privacy Sandbox. We previously discussed this here. The Trade Desk is a solid resource on discussing the downside to deprecating cookies, as they are one of Google’s fiercest opponents in that regard. The Trade Desk has famously created an alternative for a consortium of first parties and third parties called Unified Ad ID or UID2.0. There is a long and impressive list of collaborators that have adopted UID2.0.

So, who better to turn to than Jeff Green for the most recent, unfiltered commentary from one of Google’s opponents. Here is what was stated on the most recent earnings call:

Brian Fitzgerald (Analyst)

Thanks. Jeff, it looks like third-party cookies won't be going away now for at least until '25. What are your thoughts on the cookie deprecation delay once again and how, if at all, it impacts the industry? And then, maybe secondarily, could you — could the continued delays have any impacts, positive or negative on UID2.0 adoption?

Jeff Green (CEO, The Trade Desk)

[…] I think it is a strategic mistake for Google to deprecate cookies. I don't think the risk/reward is worth it for them. And I would not be surprised to see them delay this again and again as they continue to buy more time. I think that's exactly what we saw because we weren't surprised by this, we predicted this. We have just been sort of quick to move on. I do want to give Google a little bit of credit though. I mean, Apple took away cookies and said nothing, gave no announcement, offered no alternatives. Google said we're going to take away cookies, they gave some head start. Now, they moved the date a bunch, both forward and backward, which to me didn't make any sense. But they did at least try to propose something else, which was a privacy sandbox. The unfortunate thing was what they proposed was half-fake and not a valid solution. And so, the industry has just been criticizing it, us included, for the better part of the year because those criticisms, I think, were pretty unanimous even from industry bodies like the IAB that I never expected to take such a strong position on privacy sandbox. I think it forced Google's hand to delay cookie deprecation. So we were not surprised by it. The net effect of that is that it gives the open Internet a bit more runway to adopt things like UID2 and come up with authentication and identity strategies so that they can thrive in an environment outside of cookies. […]

And so, I think the average publisher is saying exactly what we were saying four or five years ago that while we think it's a strategic mistake for Google to get rid of cookies, we also think it's a strategic mistake for all the rest of us to do nothing.”

–End Quote

Google’s Antitrust Lawsuit to be Decided in Q3

We’ve covered the antitrust lawsuit in the past here when we stated: “Therein lies the issue. Google undisputedly has the world’s best consumer data, but did this grow to become part and parcel with operating a monopoly? The Department of Justice has asserted anti-trust violations against Google with the trial beginning in September 2023 […] This is not a headline to simply dismiss. It’s the first time the DOJ has brought a case of this kind against a technology company since Microsoft. If there are even minor cracks in Google’s monopoly, there could stand to be a stock or two that starts a new trajectory.”

Of course, the stock that starts the new trajectory could be Google if the DOJ rules in their favour. The trial is now concluded, and in his closing arguments, U.S. District Judge Amit Mehta questioned whether another company could rival Google’s search due to the cash and data that the company possesses, especially with Google’s 90% market share as a search engine in the crosshairs. The company pays $20 billion to make its search engine the default across Apple and Mozilla browsers and devices. Google argues they are widely used because of large R&D investments, which has made their technology superior. In cases where other search engines were the default, users complained or manually switched on their own.

The decision is expected to be announced in late summer or early fall.

Conclusion

Alphabet delivered a perfect earnings report. It beat the top line and bottom line; plus important key metrics are accelerating. Google Cloud revenue accelerated for the second consecutive quarter, narrowing the gap with Microsoft Azure. The dividend announcement was welcomed to help Google meet pressures from the Street.

Meanwhile, Alphabet has a tough year ahead with pushback from the deprecation of cookies in the early part of 2025 (assuming there are no further delays), and the looming decision from the DOJ on the antitrust trial. We continue to believe this is a landmark case for a tech company, and requires a bit of a gamble on how it will turn out for Google.

We’ve covered Alphabet extensively on AI, as well. You can find previous analysis here:

Highlights from Google I/O 2023

Posted on May 18, 2023June 30, 2026 by io-fund

Google recently held its annual developer conference Google I/O 2023. Google is a large real estate owner with arguably more data than any other tech company in the world. This advantage cannot be overstated when it comes to training large language models (LLMs). In addition to having a strategic advantage for future development of LLMs with data, Google can offer advertisers instant ROI.

The primary announcements from the event were:

Google drops the waitlist for Bard and announces new features.
Google launches new Large Language Model, PaLM2
Unveils its new AI-powered Search.
Google Cloud announces new A3 supercomputer VMs built to power LLMs.

Google drops the waitlist for Bard and announces new features

Among the more exciting announcements at Google I/O, the company dropped the waitlist for Bard and the chatbot is now available in 180 countries and territories. Bard supports English, Japanese, & Korean languages, and will soon support more than 40 languages. Google is also rolling out features such as better source citations, the ability to export content generated in Gmail and Google Docs, support for more visuals and an upcoming Google Lens integration to analyze pictures and write captions.

Background on Google’s Bard:

Earlier this year, Google’s stock (Alphabet) tumbled 7% when chatbot Bard was unable to complete a search with 100% accuracy. During the demonstration, Bard returned incorrect information about which telescope was the first to take pictures of a planet outside the Earth’s solar system. This was a minor mistake given how far large language models and generative AI has come, rather it was the timing that was a bit flawed as OpenAI’s ChatGPT, the chatbot powering competitor Microsoft Bing, had been dominating headlines since its November 30th launch.

Microsoft, being an opportunist, took it a step further and announced Bing would now be powered by a faster and more accurate version of GPT-3.5 one day after Bard’s failed demonstration: “We’re excited to announce the new Bing is running on a new, next-generation OpenAI large language model that is more powerful than ChatGPT and customized specifically for search. It takes key learnings and advancements from ChatGPT and GPT-3.5 – and it is even faster, more accurate and more capable.”

Both companies have been preparing for this moment for many years. Microsoft invested $1 billion into OpenAI a few years ago with a new $10 billion round announced last month. Meanwhile, Google acquired DeepMind in 2014. Google also previously developed conversational neural language models such as LaMDA, which is used by Google’s Bard for its conversational AI technology.

Despite the mishap with Bard, it would be a human-generated mistake to think Alphabet does not command a place of leadership right now in generative AI. Alphabet was one of the first tech companies to focus and invest on AI and natural language processing (NLP). We pointed out to our premium research members in July of 2022 that ChatGPT is based on transformer architecture that Google initially introduced in 2017 when we said:

“Transformers are becoming one of the most popular neural-network models by applying self-attention to detect how data elements in a series influence and depend on one another.

Sequential text, images and video data are used for self-supervised learning and pattern recognition, which results in more data being used to create better models. Prior to transformer models, labeled datasets had to be used to train neural networks.

Transformer models eliminate this need by finding patterns between elements mathematically, which substantially opens up what datasets can be used and how quickly.

Google first introduced transformer models in 2017 and transformers are used in Google and Bing Search. Transformers also led to BERT models, which stands for Bidirectional Encoder Representations from Transformers, and is commonly used for text sequences. Transformers are also used in GPT-3 (it’s the T in GPT) which improved from 1.5 billion parameters to 175 billion parameters. GPT-3 has the ability to report on queries it has not been specifically trained on.”

Earlier this month, Google’s CEO, Sundar Pichai, gently reminded the AI community of how cutting edge Google’s research is when he stated, “Transformer research project and our field-defining paper in 2017, as well as our important advances in diffusion models, are now the basis of many of the generative AI applications you're starting to see today.”

BERT was designed to help Google better understand search intent, as despite billions of searches every day, about 15% of those searches are for brand new terms. This prompted Google engineers to develop a model that could self-learn.

The result is that searches results are more accurate by taking into consideration the nuances of language.

Google launches new Large Language Model, PaLM2

Google launched a new Large Language Model, PaLM2, that will power the updated Bard AI chat tool and more than 25 other new products & features including productivity software (Gmail, Google Docs), Healthcare and Security.

PaLM 2 has the following capabilities:

Multilingual: The LLM is trained on more than 100 languages, which increases language proficiency
Reasoning: The LLM’s dataset has improved logic, common sense reasoning and mathematics
Coding: The LLM can generate code including programing languages such as Python, JavaScript and specialized languages such as Prolog, Fortran and Verilog.

Google Unveils its new AI-powered Search

The company has unveiled its new generative AI-powered search that will be subject to a waitlist. Google cites the example of the following search “what's better for a family with kids under 3 and a dog, bryce canyon or arches.” Previously, you had to break the question down into smaller ones, sort through the vast information available, and then put things together yourself. Now with generative AI, search will be able to better understand the question.

Generative AI will also provide a better experience for online shopping by instantly getting relevant information like reviews, images, and ratings. The new shopping experience is based on Google’s Shopping Graph, which has more than 35 billion product listings.

The company announced the ‘About this image’ feature, allowing users to identify fake images. It mentioned in its press release, “When the image and similar images were first indexed by Google, Where it may have first appeared, and Where else it’s been seen online (like on news, social, or fact checking sites)”.

Google launches new Large Language Model, PaLM2

The company launched the new Large Language Model, PaLM2, that will power the updated Bard AI chat tool and more than 25 other new products & features announced during the Google I/O 2023.

Its predecessor PaLM, launched in April 2022, was a 540 billion based parameter, and the company did not provide this detail for PaLM2. PaLM stands for Pathways Language Model. “What we found in our work is that it’s not really the sort of size of model — that the larger is not always better,” DeepMind VP Zoubin Ghahramani said in a press briefing ahead of the announcement. “That’s why we’ve provided a family of models of different sizes. We think that actually parameter count is not really a useful way of thinking about the capabilities of models and capabilities are really to be judged by people using the models and finding out whether they’re useful in the tests that they try to achieve with these models.”

PaLM2 is faster and more efficient than previous models. Some of the improvements highlighted by the company are that PaLM2 is trained for improved multilingual text, spanning over 100 languages, reasoning, and coding, including popular languages like Python & JavaScript. For example, due to the multilingual capabilities of PaLM2, it has helped Bard to expand to new languages. PaLM2 is available in four sizes: Gecko, the smallest, followed by Otter, Bison, and Unicorn. Other use cases include improved Workspace features while working in Gmail, Google Docs, and Google Sheets. PaLM2 can also be used for enterprise use cases like Med-PaLM2 in medical research and Sec-PaLM in cybersecurity.

The company also said that it’s working on a more powerful model called Gemini and it will also be available in various sizes so that it can be easily deployed to various products.

Google Cloud announces new A3 supercomputer VMs built to power LLMs

Google Cloud announced the A3 GPU supercomputer that can be used to train and run Artificial Intelligence and Machine Learning models. While the A3 GPU supercomputer is on a private preview waitlist, the previously announced G2 VMs are now in general availability. The G2 VMs are powered by the new Nvidia L4 Tensor Core GPUs. The company said that it is the first cloud provider to offer these new GPUs for serving generative AI workloads.

The A3 GPU VMs are made of eight Nvidia H100 Hopper architecture GPUs, 3.6 TB/s bisectional bandwidth between A3’s 8 GPUs via the Nvidia NVSwitch and NVLink 4.0, 4th Gen Intel Xeon Scalable processors, and 2TB of host memory.

The A3 supercomputer can deliver up to 26 exaFlops of AI performance, thereby improving the time and cost of training large machine learning models. The A3 workloads will be run on Google’s Jupiter data center networking fabric which the company states “scales to tens of thousands of highly interconnected GPUs and allows for full-bandwidth reconfigurable optical links that can adjust the topology on demand.”

Conclusion:

I would not be surprised if we exit 2023 with a reimagined way to use Search Engines. The iteration cycle here is likely to move quickly compared to AVs or the Metaverse, as there are real-world applications where AI can be applied without safety issues (AVs) or friction in terms of user adoption (Metaverse/VR headsets). Instead, the scale has already been built with Search being a viral, daily activity used by nearly every human on earth. AI advancements will simply improve what is already in place.

Cutting-edge chatbots can be quickly deployed on the search engines that already exist, and this is a substantial difference from other overhyped, early-stage technologies. Their accuracy may still need time, but they're probably not too far off from being deemed “reliable enough.”

Investors should expect that AI will become a winner(s)-take-all market. In time, the difference in how search and other applications operate in terms of user experience plus ROI for advertisers will help carve a larger lead.

Premium Members should check the forum for updates on our timing for an entry into the stock.

Highlights from Google I/O 2023

Posted on May 18, 2023June 30, 2026 by io-fund

The primary announcements from the event were:

Google drops the waitlist for Bard and announces new features.
Google launches new Large Language Model, PaLM2
Unveils its new AI-powered Search.
Google Cloud announces new A3 supercomputer VMs built to power LLMs.

Google drops the waitlist for Bard and announces new features

Background on Google’s Bard:

“Transformers are becoming one of the most popular neural-network models by applying self-attention to detect how data elements in a series influence and depend on one another.

Transformer models eliminate this need by finding patterns between elements mathematically, which substantially opens up what datasets can be used and how quickly.

The result is that searches results are more accurate by taking into consideration the nuances of language.

Google launches new Large Language Model, PaLM2

PaLM 2 has the following capabilities:

Multilingual: The LLM is trained on more than 100 languages, which increases language proficiency
Reasoning: The LLM’s dataset has improved logic, common sense reasoning and mathematics
Coding: The LLM can generate code including programing languages such as Python, JavaScript and specialized languages such as Prolog, Fortran and Verilog.

Google Unveils its new AI-powered Search

Google launches new Large Language Model, PaLM2

The company launched the new Large Language Model, PaLM2, that will power the updated Bard AI chat tool and more than 25 other new products & features announced during the Google I/O 2023.

The company also said that it’s working on a more powerful model called Gemini and it will also be available in various sizes so that it can be easily deployed to various products.

Google Cloud announces new A3 supercomputer VMs built to power LLMs

Conclusion:

Premium Members should check the forum for updates on our timing for an entry into the stock.