Signals Inbox·August 30, 2026·Frontier AI

AI Race: who is losing it badly as of today?

Apple is losing the Big Tech AI race by the widest margin right now. Google looks strongest overall, Microsoft has the clearest commercial proof, and Meta has recovered enough technically that it no longer belongs in the loser category.

Q1Why do people keep comparing Google, Microsoft, Meta and Apple in AI?

Google, Microsoft, Meta and Apple all control huge distribution channels that can turn AI into a default habit for billions of people.

Google has Search, Android, Gmail, YouTube, Workspace and Cloud. Microsoft has Windows, Microsoft 365, Azure and GitHub. Meta has Facebook, Instagram, WhatsApp and Messenger. Apple controls the iPhone, Mac and the operating systems sitting directly between users and their devices.

That puts the four companies in a fairly specific race: who can become the default AI layer inside products people already use every day?

OpenAI and Anthropic belong in the wider competitive picture, but their starting point is different. Google, Microsoft, Meta and Apple already owned massive platforms before the current AI boom. Their challenge is turning that installed distribution into actual AI usage without letting a new AI-native company take control of the interface.

The comparison has become even more interesting lately because all four have now shown much more of their hand. Google has pushed Gemini deeper into Search and crossed a major adoption milestone. Microsoft has disclosed large paid Copilot deployments. Meta has returned with a much stronger generation of models. Apple has finally unveiled the Siri AI it had been promising.

Q2Why is it so hard to say which Big Tech company is winning AI?

Google, Microsoft, Meta and Apple are leading on very different scoreboards right now.

Google has the broadest combination of consumer usage, models, search distribution, developers and cloud growth. Microsoft has the strongest evidence that companies are actually paying for AI at scale. Meta has recently become much harder to dismiss on model quality. Apple has the most powerful device distribution of the four, yet much less proof that people are already using its new AI layer.

Revenue alone gives us a poor answer. Model benchmarks do too. Installed base can also be misleading when users have access to the technology but rarely use it.

The useful scoreboards are actual usage, paid enterprise adoption, developer adoption, model quality, product speed, control of infrastructure and whether AI is strengthening the existing business. They point in different directions, but put together they give us a fairly clear ranking.

Q3Who has the biggest real consumer AI audience right now?

Google currently has the strongest measured consumer AI reach among these four companies.

Google recently said the standalone Gemini app had crossed 1 billion monthly users, up from 950 million around its previous earnings report. Google also says AI Mode inside Search has passed 1 billion monthly users. Those audiences overlap, so adding the two numbers would make no sense, but each one is huge on its own. Gemini daily users have also tripled over the past year.

Meta has even more raw distribution. Its latest results show 3.60 billion people using its family of apps daily. The problem is that a Facebook, Instagram or WhatsApp user cannot automatically be counted as a Meta AI user. Meta currently gives us much less clean disclosure around how often those billions of people deliberately use its AI assistant.

Microsoft's strongest usage evidence comes from work rather than the mass consumer market. Apple, meanwhile, has no current Siri AI usage figure that belongs in the same conversation.

Google wins this part of the race by a wide margin.

Q4Which company has built the strongest AI business so far?

Microsoft currently has the strongest paid AI machine, while Google has the hotter cloud growth.

Microsoft's latest quarter showed Azure growth of 43%, with demand still exceeding available capacity. Microsoft Cloud generated $59.3 billion in quarterly revenue, although that figure includes plenty of services unrelated to generative AI. GitHub Copilot revenue also accelerated more than 60% quarter over quarter after Microsoft changed its pricing model.

Google is growing considerably faster from a smaller cloud base. Google Cloud revenue jumped 82% year over year, helped by AI infrastructure and AI products, while backlog reached $514 billion. Google's reported cloud growth rate is roughly twice Azure's current rate, although the two companies define their cloud businesses differently.

Neither company gives us a clean line called "AI revenue," so an exact comparison would be fake precision. The commercial evidence still favors Microsoft today because it has already embedded AI into enormous paid software and cloud businesses. Google is closing quickly.

Meta and Apple are much further back on direct AI monetization.

Q5Which company has the strongest proof that big companies really use its AI?

Microsoft has the best evidence that enterprise AI has moved beyond experiments.

Microsoft says Microsoft 365 Copilot has passed 30 million paid seats. More importantly, customers with over 50,000 seats increased more than sevenfold year over year, while the number of enterprises deploying Copilot to most of their information workers jumped nearly 75% in a single quarter. HSBC committed to 200,000 seats, KPMG is expanding across more than 276,000 employees, and NHS England is rolling it out to 505,000 clinicians and staff.

Google's enterprise numbers are also serious. Nearly 90% of the Fortune 100 now use Gemini Enterprise, according to Google's latest earnings call. Almost 500 Cloud customers processed more than one trillion tokens each over the past year, and existing customers are exceeding their original commitments by more than 50%.

The difference is evidence quality. "Using Gemini Enterprise" covers a wide range of deployment sizes. Microsoft's seat counts tell us exactly how many licenses companies have bought and show deployments spreading across entire workforces.

Microsoft is ahead here.

Q6Who has the strongest hold on AI developers?

Microsoft currently has the strongest developer distribution, with Google close behind.

GitHub Copilot now has 50 million users, while GitHub itself has reached 225 million users. Microsoft also says one in three pull requests on GitHub now involves an agent. That puts AI directly inside one of the places where software developers already spend their day.

Google approaches developers through models and infrastructure. More than 9 million developers each month are building with Google's models and developer products, and Google's model APIs are processing roughly 22 billion tokens per minute, up from 16 billion one quarter earlier.

Those numbers measure different things, so there is no sensible ratio between them. Microsoft has the cleaner distribution advantage because GitHub owns the workflow itself. Google has an extremely strong model and cloud ecosystem sitting right behind it.

Meta has only recently opened its new Meta Model API around Muse, while Apple's Foundation Models framework remains much earlier as a developer platform.

Q7Whose AI models are actually strongest right now?

Meta has the strongest fresh public model result among these four companies.

The latest overall Arena text leaderboard places Muse Spark 1.2 fourth overall. Meta ranks second among labs on that leaderboard, behind Anthropic. Google's listed Gemini 3.7 Flash High model sits lower in the current ranking and is still marked preliminary.

That is a much bigger deal for Meta than one benchmark position might suggest. Meta spent a long period looking like it had fallen behind the frontier. Muse Spark 1.2 gives us recent independent evidence that Meta can again produce a model people prefer against many of the strongest alternatives.

Arena still has limitations. Muse Spark 1.2 has fewer votes than several mature Claude models, rankings move quickly, and model quality varies by task. Meta's own latest work also focuses heavily on coding, agentic behavior and multimodal reasoning rather than winning every benchmark.

Still, calling Meta technically lost today would be hard to defend.

Q8Is Meta still one of the companies falling badly behind?

Meta is currently behind Google and Microsoft commercially, but it has pulled itself back into the technical race.

The recent sequence is hard to ignore. Meta introduced Muse Spark, followed it with Muse Spark 1.1 and a developer API, added agentic features that can work with email and calendars, then released Muse Spark 1.2 and Muse Code. Meta has since shown stronger multimodal capabilities and released Muse Glimmer, an agentic model designed to run on devices.

That looks much healthier than a company depending on one flashy launch.

Meta's weak spot these days is economic proof. The company can use AI to make Instagram, Facebook and its advertising systems better, so generative AI does not necessarily need its own revenue line. Even so, Google and Microsoft already show obvious paid demand for AI infrastructure, software and developer tools. Meta still has much less evidence there.

Meta now looks like a company trying to turn a technical recovery into a business, rather than one watching the frontier disappear.

Q9Which Big Tech company is shipping useful AI the fastest?

Google currently has the strongest product velocity across the widest number of products.

Google is pushing AI through Search, the Gemini app, Workspace, Cloud, Android and YouTube at the same time. AI Mode has already reached mass-market usage, while Ask YouTube reached more than 140 million users on the watch page during Google's latest reported quarter. Google has also kept releasing new Gemini variants while preparing the next generation of its models.

Microsoft is moving extremely quickly around Copilot, GitHub and enterprise agents. Meta has probably shown the biggest recent acceleration, especially after the rapid Muse releases.

Apple is the obvious fourth place today. It unveiled a dramatically more capable Siri AI, but the new Siri is currently available for developer testing, with a broader user beta coming later. There is still a big gap between what Apple has shown and what hundreds of millions of users can already use from Google, Microsoft and Meta.

Q10Which company controls the whole AI stack best?

Google controls the most complete AI stack among these four companies.

Google builds frontier models through DeepMind, designs its own TPUs, runs Google Cloud, sells model access to developers, distributes Gemini directly to consumers and inserts the same technology into Search, Android, Workspace and YouTube.

That setup gives Google unusual control over both cost and distribution. Google says it has already pushed the cost of serving AI Mode to its lowest level since launch even while adding more capable models.

Microsoft comes close through Azure, its Maia chips, GitHub, Microsoft 365 and its own model work. Microsoft has also deliberately built a multi-model ecosystem around OpenAI, Anthropic and others. Customers get flexibility, although Microsoft depends more heavily on outside frontier labs than Google does.

Meta controls models, apps and a rapidly growing infrastructure footprint, though it lacks an external cloud business comparable with Azure or Google Cloud.

Apple has excellent control over devices and silicon. Its model layer is less independent: Apple's third-generation Foundation Models were custom-built in collaboration with Google, according to Apple's own machine-learning team.

Google has the cleanest end-to-end position.

Q11Which company is using AI to make its old business stronger?

Google currently has the clearest proof that AI is strengthening the business people once thought it might destroy.

Search was supposed to be one of generative AI's first major victims. Google's latest results went in the opposite direction. Search and Other revenue grew 17% year over year, while Google says AI-powered features are increasing the number of searches people make. AI Mode has already reached enormous scale without producing the collapse in search economics that many people expected.

That carries more weight than a model leaderboard. Google is changing its flagship product while the flagship product keeps growing.

Meta is also benefiting from AI inside advertising and recommendations. Its latest quarterly revenue rose 28%, while ad impressions increased 14% and average price per ad rose 12%. Meta attributes some of its core-business improvement to AI, although the financial disclosures do not isolate the exact contribution.

Google has gone further because the new AI interface is already visible to users at massive scale and the old money machine is still growing.

Q12Can Apple's 2.5 billion devices still save it in the AI race?

Apple's enormous installed base can still rescue its AI position, but it has not rescued it yet.

Apple has more than 2.5 billion active devices. Once Apple has an AI assistant people genuinely love, distribution becomes much easier. Siri sits inside the operating system, can see personal context, interact with apps and eventually work across a user's messages, emails, photos and screen. Apple does not have to convince an iPhone owner to discover a completely new platform first.

For now, the usage proof is missing. Siri AI remains in developer testing, with Apple planning a broader beta later. Apple's own third-generation AI models were also built with Google, which makes Apple's current position unusual for a company that normally likes to own its core technology.

Apple can afford to fix this. Its latest quarter produced $109.4 billion of revenue, up 16% year over year, and its active-device base reached another record. Apple's underlying business is doing extremely well.

The problem is time. Google, Microsoft and Meta are building usage habits while Apple is still getting its new assistant into users' hands.

Q13Which company is taking the biggest financial gamble on AI?

Meta is taking the riskiest AI bet relative to how little standalone AI revenue it currently discloses, even though Google and Microsoft are spending more in absolute dollars.

Meta spent $31.1 billion on capital expenditures in its latest quarter and expects full-year spending of $130 billion to $145 billion. Quarterly free cash flow fell to $784 million during that heavy investment period. Meta still had $90.3 billion in cash and marketable securities, so funding the race is hardly an immediate problem.

What makes the bet unusual is the revenue mix. Of Meta's $60.8 billion of quarterly revenue, $59.4 billion came from advertising. AI helps that advertising machine, but Meta has yet to show anything comparable to Azure, Google Cloud or Microsoft 365 Copilot as a separately visible commercial AI engine.

Google expects around $180 billion to $190 billion of capital expenditure this year, while Microsoft now expects roughly $175 billion after an accounting-related lease reclassification. Both are spending more than Meta in absolute terms, but both can point to very large cloud businesses where new AI capacity is already being sold.

Meta can absolutely make the economics work. Right now, though, it has the biggest gap between spending ambition and visible direct AI monetization.

Q14Does the wider AI market make Apple's position look even worse?

The broader AI market makes Apple look further behind, because the frontier keeps moving while Apple is still rolling out its first truly competitive Siri generation.

Anthropic currently occupies the top positions on Arena's overall model ranking. OpenAI is still releasing upgraded GPT-5.6 models and expanding ChatGPT's capabilities. Meta has returned near the top of public model rankings. Google keeps improving Gemini while distributing it through products with billions of users.

Microsoft has a useful hedge against this constant model turnover. Azure and Foundry can sell access to multiple model families, so Microsoft can benefit even when the world's best model comes from another lab. Foundry now has 100,000 customers, according to Microsoft's latest earnings call.

Apple currently depends more heavily on bringing outside intelligence into its own hardware and software experience. That strategy could eventually work extremely well, especially if frontier models become cheap and interchangeable.

Today, owning the user experience without controlling the fastest-moving intelligence layer leaves Apple with less strategic control than Google and Meta. Microsoft, meanwhile, has built a business around selling access to whoever wins the model race.

Q15So who is losing the Big Tech AI race badly right now?

Apple is clearly last among Google, Microsoft, Meta and Apple today.

Google looks strongest overall. Microsoft sits very close behind and is ahead in some of the areas where actual money is already changing hands. Meta is third, with a technical recovery strong enough that calling it a major loser no longer fits the evidence. Apple is fourth.

Big Tech AI race by competitive dimension

Three parts of this comparison carry more weight than the rest.

The first is real usage. Google already has consumer AI at huge scale, while Apple's newest Siri experience has yet to reach ordinary users broadly.

The second is commercial proof. Microsoft can show paid Copilot seats, huge deployments and cloud capacity being sold as quickly as it comes online. Google can show explosive Cloud growth and very heavy enterprise model usage. Meta's commercial story is thinner, and Apple's is thinner again.

The third is control of the technology. Google owns almost every major layer needed to compete. Meta has regained credibility at the model layer. Microsoft can combine its own technology with whichever outside models customers want. Apple's collaboration with Google may eventually be a smart shortcut, but right now it reinforces the impression that Apple is catching up rather than setting the pace.

So our ranking today is Google first, Microsoft second, Meta third and Apple fourth.

Apple still has the easiest path to making that ranking look foolish. A genuinely excellent Siri AI can arrive on an installed base measured in billions and become a major product almost overnight. What Apple needs to prove over the next several quarters is that Siri AI moves from a promising demo and developer beta into something people actually use constantly.

Meta needs to show that its recent technical comeback can produce meaningful AI revenue or a large new business inside its existing apps. Microsoft needs to keep turning Copilot deployments into sustained usage as the easy wave of corporate pilots passes. Google needs to show that Gemini and AI Search can keep growing without weakening the search economics that finance the whole operation.

Until those facts change, the answer is fairly hard to dodge: Apple is losing the Big Tech AI race by the widest margin right now, while Google is winning the broadest version of it.