Public wars often come with a hefty price tag. While OpenAI and Anthropic are caught up in legal and public relations battles that consume media oxygen, Google has been quietly crafting a devastating strategy. It’s growing its user base faster than its rivals while simultaneously securing military contracts that other competitors can only dream of. Ironically, the labs that promised to democratize AI are ceding ground to the giant they vowed to dethrone.
The numbers don’t lie. Gemini surpassed 300 million monthly active users in February 2026, doubling its user base in just six months. Meanwhile, ChatGPT, despite its first-mover advantage, hovers around 250 million. Claude, despite its technical excellence, barely reaches 80 million. But what’s truly revealing isn’t just the consumption numbers; it’s Google’s quiet advance in strategic territory: government contracts, critical infrastructure, and especially the U.S. Department of Defense.
The Pentagon Chooses a Winning Horse (and It’s Not the One You Expected)
Google Cloud has just renewed and expanded its contract with the DoD. This agreement incorporates Gemini capabilities specifically designed for intelligence analysis and logistical operations. Valued at $2.3 billion through 2029, the contract includes exclusivity clauses for certain segments of classified information processing.
This represents a 180-degree turnaround from 2018 when Google faced an internal revolt over Project Maven and had to withdraw from military contracts. What changed? Two key things: first, Google redefined its "AI Principles" with enough ambiguity to work with defense without publicly violating its commitments. Second, and more importantly, its competitors handed it the opportunity.
While Anthropic got tangled in public disputes over its Constitutional AI, and OpenAI grappled with internal leaks about its dual-use technology, Google simply executed. No fanfare, no grandiose promises about "alignment" or "safety." Just solid engineering and institutional relationships cultivated since the Eric Schmidt days.
The Paradox of Ideological Purity
OpenAI promised never to work with offensive military applications. Anthropic was explicitly founded on principles of "safe AI." Both positions are admirable but commercially naive. The U.S. government spends over $3.5 billion annually on AI and machine learning contracts. Excluding oneself from that market doesn’t make it disappear; it simply ensures someone else will capture it.
Google learned the lesson from Maven: don’t reject government contracts; redefine them. Its systems for the DoD are officially categorized as "defensive" and "logistical support." The line is blurry, of course, but clear enough for its internal stakeholders and vague enough for its ethical auditors.
Silent Speed: How Gemini Won the Adoption Race
Google's distribution strategy is brutally simple: be where users already are. Gemini doesn’t need to convince you to create a new account, change your workflow, or learn a new interface. It simply appears in Gmail, Docs, Search, and Maps. By March 2026, 73% of Gemini users never visited gemini.google.com; they accessed it through products they already use daily.
ChatGPT has to win over every user. Gemini just has to not annoy them too much. It’s an architectural advantage that, honestly, no independent lab can replicate.
The Business Model No One Wants to Admit
OpenAI sells subscriptions. Anthropic sells API calls. Google, on the other hand, gives away AI to sell you more storage, more ads, and more cloud infrastructure. Its incentives are perfectly aligned with mass adoption, even at an initial loss. They can subsidize Gemini indefinitely because every Gemini user is a deeper user of the Google ecosystem.
In the first quarter of 2026, Google Cloud grew 34% year-over-year, primarily driven by AI workloads. It’s not just about Gemini inference loads; it’s about companies migrating their entire infrastructure to take advantage of integration with Vertex AI, optimized TPUs, and yes, embedded Gemini models.
Claude may be technically superior in long-form reasoning (it is). GPT-4.5 may be more creative in certain tasks (debatable). But Gemini is everywhere, functioning "well enough" for 80% of use cases. And in tech, ubiquity often trumps excellence.
The Wrong War at the Wrong Time
OpenAI and Anthropic are caught in a narrative of differentiation that is losing relevance. They argue about Constitutional AI vs. RLHF, about "alignment taxes" and fine-tuning methodologies, while the business world has come to a simpler conclusion. Any frontier model is good enough for most productive applications.
The public battles between Sam Altman and Dario Amodei over AI safety approaches sound increasingly academic to a CFO who just wants to automate their customer support department. And that CFO likely already has Google Workspace, making the implementation of Gemini a checkbox issue, not a project.
The Cost of Fragmented Attention
Every hour that OpenAI spends responding to controversies about its corporate structure, Google gains 50,000 new Gemini users. Every article about internal disputes at Anthropic is media oxygen that isn’t spent discussing the limitations of Gemini or the risks of concentration in Google.
Fragmented attention always favors the incumbent. Google doesn’t need to win the innovation narrative; it just needs its competitors to cannibalize each other's mindshare while it quietly executes.
Why This Matters Beyond the Valley
In 2026, we are witnessing the potential premature consolidation of a market that promised to be plural. If Google simultaneously captures the government/defense segment, the enterprise market through Cloud, and mass consumer access via product integration, what space is left for independent labs?
OpenAI may raise another $10 billion. Anthropic may remain the favorite of purist developers. But neither has the 14 products that exceed a billion users that Google does. Nor do they have the institutional relationships Google has built over 25 years. Definitely, neither can operate with the negative margins that Google tolerates in its AI division, subsidized by $300 billion in annual revenue from other lines.
The question for founders and implementers isn’t "which model is better?" but "are we betting our infrastructure on a provider that might not be around in three years?" And there, Google’s corporate solidity starts to look less boring and more strategic.
Is it possible that diversity in AI models is a temporary luxury that the market cannot sustain? As you read this, 3,000 more companies have just activated Gemini in their workspace. The war already has a winner; it’s just a matter of someone admitting it.