Why Google Backs Claude While Building Gemini

At first glance, Google supporting Anthropic while also pushing Gemini looks contradictory. In practice, it is easier to understand if you separate four things: investment upside, infrastructure revenue, enterprise demand, and product benchmarking.

Google’s in-house model family, Gemini, competes with Anthropic’s family of AI models, Claude. But Google can still benefit if Claude grows, because Anthropic is also an infrastructure customer, a distribution partner on Google Cloud, and a company in which Google holds an economic interest.

The short version is simple: Google does not need Claude and Gemini to be friends. It needs exposure to more than one path to value in a fast-moving AI market.

Quick answer: Why Google backs Claude while building Gemini

Google can benefit from Anthropic even while Gemini competes with Claude. It can win as an investor, as a cloud and compute provider, and as a company serving enterprise buyers who may want access to more than one frontier model.

That is the hedge thesis in plain English: if Gemini becomes the main winner, Google benefits directly. If Claude keeps gaining traction, Google can still benefit through its stake, its infrastructure relationship, and its cloud distribution.

A simple analogy is a company that sells tools to a rival while also competing in the finished product market. The rivalry is real, but the supplier still gets paid. The rest of this article breaks down the strategy through four lenses: hedge logic, cloud revenue, enterprise demand, and benchmarking.

Definition box: Anthropic, Claude, Gemini, and what “backing” really means

If you want more background, see our guide to What is Anthropic and who funds it.

Why this looks contradictory at first

The confusion is fair. Claude and Gemini compete for developers, enterprise budgets, and mindshare. Google is also investing heavily in its own model stack, so funding or supporting a rival can look strange from the outside.

But large platform companies often compete at one layer and cooperate at another. In this case, Google and Anthropic have both a horizontal relationship, because Gemini and Claude compete, and a vertical relationship, because Google supplies Anthropic with compute and distribution services.

A useful example is cloud software more broadly: a cloud provider might host a competing application company while also selling its own app. The application rivalry remains real, but the infrastructure provider still has a reason to support customer demand.

So the relationship is neither purely friendly nor purely adversarial. It is best understood as strategic overlap with built-in tension.

Reason 1: Google is hedging its bets in a fast-moving AI market

The first reason is portfolio logic. Frontier AI is evolving quickly, and no one can assume a single model family will dominate every workload, buyer segment, or regulatory environment. Backing Anthropic gives Google exposure to another possible winner while it continues to push Gemini internally.

That does not mean Google is giving up on Gemini. A hedge is not surrender. It is a way to reduce concentration risk when the market could fragment across several leading models.

In practical terms, Google can benefit under multiple outcomes. If Gemini wins broad adoption, Google captures the upside directly through its own products and ecosystem. If Claude takes share in key enterprise or developer use cases, Google still has economic and infrastructure exposure to that growth.

This layered logic also matches how regulators have described the relationship. The UK CMA characterized Google and Anthropic as both competitors and partners, while the FTC included Google-Anthropic among the major cloud-provider and AI-developer partnerships it examined.

One plain-language example: imagine you build your own electric car brand but also invest in a battery supplier that serves the wider market. You still want your own car to win, but you have positioned yourself to benefit if demand expands in other directions too.

The key caution is that “hedge” is an analytical framing. It is a useful way to describe the strategy, but readers should treat exact investment narratives and totals as dependent on official disclosures and reporting updates.

Reason 2: Google Cloud and TPU capacity can benefit from Anthropic’s growth

The second reason is infrastructure economics. In AI, the value does not sit only in the model layer. It also sits in cloud usage, accelerator demand, hosting, inference, governance tooling, and enterprise distribution.

That is where cash-and-compute relationships matter. In simple terms, a cloud provider can support an AI developer through funding, hardware access, cloud capacity, or a combination of those. The FTC staff report says these partnerships generally include cloud-computing spending commitments, and some include public exclusivity requirements.

Anthropic has publicly said it is significantly expanding its use of Google Cloud TPUs and services. Google also said Anthropic was one of the first companies to deploy TPU v5e at scale to serve Claude.

For Google, that means Anthropic’s growth can translate into infrastructure demand. Even if Anthropic owns the model layer, Google can still earn from the compute layer. That is a different revenue stream from selling Gemini itself, but it is still strategically valuable.

A simple example helps. If an enterprise runs Claude through Google Cloud, Google may not own the model IP, but it still participates through the cloud relationship, platform tooling, and infrastructure consumption. That is why infrastructure revenue and model revenue should be treated separately.

This also improves Google Cloud’s position in enterprise AI. If buyers want optionality across model vendors, hosting Claude on Google Cloud makes the platform more attractive. Enterprises often prefer not to lock themselves into a single model family, especially while the market is still changing quickly.

Anthropic also says Claude is available on Google Cloud, including through Vertex AI, with Google Cloud infrastructure and governance capabilities. That makes the relationship commercially meaningful, not just financially symbolic.

If you want the mechanics behind these arrangements, read our explainer on How AI cloud compute deals work.

Reason 3: Enterprise customers may prefer Claude in some workflows, and Google cannot ignore that

The third reason is customer demand. Enterprise AI is not only about public benchmark bragging rights. It is about recurring contracts, workflow integration, procurement confidence, security review, and whether teams actually keep using the model later.

Anthropic positions Claude on Google Cloud for enterprise deployment, emphasizing governance, security controls, and scalable infrastructure. It also highlights use cases such as AI agents, coding, customer support, and financial analysis.

That does not prove Claude is universally better than Gemini. But it does show why Google has reason to support customer choice when enterprise buyers want access to Anthropic’s models in a Google Cloud environment.

In the market, Claude is often described as competitive in enterprise-oriented and long-document workflows. The safest takeaway is not that one model always wins. It is that enterprise buyers frequently evaluate models by task fit, reliability, context handling, safety behavior, and procurement comfort, not just by headline marketing.

Anthropic also said its customers ranged from large enterprises to AI-native startups and that expanded compute capacity would help it meet growing customer demand. If enterprise demand is real, Google benefits from meeting customers where they are instead of forcing a single-model story.

For a buyer, this is practical. If your teams prefer Claude for a specific workflow but your cloud strategy centers on Google Cloud, having Claude available there reduces friction. That makes Google Cloud more useful as a platform even when Gemini is not the chosen model for every job.

For more on how buyers think about this, see our guide to Enterprise AI model selection criteria.

Reason 4: Competitive benchmarking can help Google improve Gemini

The fourth reason is product learning. Serious AI teams compare outputs across models all the time. That includes tests for reasoning quality, coding performance, hallucination behavior, safety refusals, and long-document handling.

That kind of comparative testing does not imply ownership or dependence. It is standard product practice. If a rival model performs better on a task that matters to your customers, you want to know.

There is also reported evidence on this point. TechCrunch reported that Google contractors comparing Gemini outputs were shown Claude responses in Google’s internal evaluation platform. TechCrunch also reported that Google said DeepMind does compare model outputs as part of its evaluation process but does not train Gemini on Anthropic models.

That distinction matters. Benchmarking against a rival is normal. Training on a rival’s model outputs would be a very different claim, and the reporting did not say that.

For enterprise readers, the practical takeaway is straightforward: model developers learn from side-by-side evaluation. That does not remove competition. It sharpens it.

How Google wins whether Claude or Gemini gains traction

The easiest way to understand the strategy is to map the two main scenarios.

If Gemini wins strongly

• Google gains direct product upside from its own model family.
• It strengthens control over its application stack, developer ecosystem, and enterprise distribution.
• It can capture model revenue more directly rather than only infrastructure revenue.

If Claude keeps gaining traction

• Google still has investment exposure through its stake and related rights.
• Google Cloud and TPU usage can benefit from Anthropic demand.
• Google stays relevant to enterprise buyers who want access to Claude through Google Cloud and Vertex AI.

The FTC staff report also notes that these kinds of partnerships can provide cloud partners with significant equity and certain revenue-sharing rights. The details matter by deal, but the broader point is clear: the cloud partner can still win even when the model developer is not the same company.

This does not remove competitive risk. It reduces concentration risk.

Claude vs Gemini: a practical comparison table for this strategy question

If you are trying to understand the strategy rather than crown a technical winner, this is the comparison that matters.

Dimension	Claude	Gemini
Who owns/builds it	Built by Anthropic, which the CMA described as developing the Claude family of foundation models.	Built by Google, which the CMA described as developing the Gemini family of foundation models.
How Google benefits	Indirectly through investment exposure, cloud and TPU demand, and enterprise distribution via Google Cloud and Vertex AI.	Directly through ownership of the model family, product integration, and ecosystem control.
Role in enterprise AI	Available on Google Cloud and Vertex AI for enterprise use cases such as agents, coding, customer support, and financial analysis.	Google’s flagship in-house model family for its own AI stack and enterprise offerings.
Infrastructure relationship	Anthropic has publicly expanded use of Google Cloud TPUs and services; the CMA described the compute supply and distribution arrangements as non-exclusive.	Runs within Google’s own infrastructure ecosystem.
Distribution	Distributed through Anthropic directly and through Google Cloud’s Vertex AI.	Distributed through Google’s own products, APIs, and cloud channels.
Strategic value to Google	Provides hedge exposure, cloud revenue opportunity, and a way to serve customers who want model choice.	Provides direct competitive positioning, product differentiation, and tighter control over roadmap and margins.
Strategic risk to Google	Supporting Claude can strengthen a direct competitor to Gemini and create channel conflict in enterprise accounts.	If Gemini underperforms in segments where buyers prefer alternatives, Google could lose model-layer share even while keeping some cloud exposure.

For a deeper buyer-oriented view, read our Claude vs Gemini comparison.

Timeline: Google’s reported Anthropic commitments and what changed over time

The timing matters because the relationship has evolved across investment, infrastructure, and distribution.

• October 2022: The CMA said the earliest iteration of the Google-Anthropic side letter was signed in October 2022.
• November 2023: Google announced an expanded AI partnership with Anthropic and said Anthropic would leverage Cloud TPU v5e chips for AI inference.
• March 2024: Anthropic announced that Claude 3 Haiku and Claude 3 Sonnet were generally available on Google Cloud’s Vertex AI platform.
• August 2024: The CMA said the most recent version of the side letter it reviewed was signed in August 2024.
• October 2024: Anthropic said it was significantly expanding its usage of Google Cloud TPUs and services.

The CMA also said the broader partnership includes non-voting Anthropic shares, convertible debt financing, a non-exclusive compute supply arrangement, and a non-exclusive distribution agreement on Vertex AI.

That is an important distinction. The relationship is not just cash. It combines financing, infrastructure access, and route-to-market. It is also important not to collapse every reported amount or structure into one simple number, because disclosures can refer to different instruments and stages over time.

What this says about the AI market

The bigger lesson is that frontier AI is a layered market. The winners in models, cloud infrastructure, accelerators, and enterprise distribution do not have to be the same companies.

That is why platform companies can support rivals at one layer while competing at another. A model provider may compete with Google in AI capabilities but still depend on Google for compute, cloud channels, or enterprise access.

You can see the same pattern across tech more broadly. Infrastructure suppliers often serve companies that also compete with them in adjacent markets. What matters is who controls which layer and where the margins are.

In that sense, the Google-Anthropic relationship is not an odd exception. It is a sign that AI strategy increasingly looks like ecosystem management, optionality, and distribution control rather than a simple winner-take-all race.

Risks and tensions in Google backing a rival model

The strategy is rational, but it is not risk-free.

The first risk is obvious: supporting Anthropic can strengthen a direct competitor to Gemini. If Claude becomes the preferred option in some enterprise accounts, Google may keep infrastructure exposure while losing higher-value model-layer revenue.

The second risk is channel conflict. Enterprise buyers may choose Claude over Gemini even when both are available through Google-related pathways. That can complicate sales narratives and make product positioning less clean.

The third risk is strategic ambiguity. Partners, customers, and observers may ask what Google really wants the market to believe. Is Gemini the centerpiece, or is Google optimizing for a multi-model platform outcome? The answer can be both, but mixed signals can still create confusion.

There is also a reputational trade-off. If Google presents Gemini as its core strategic model family while heavily supporting a rival, some buyers may read that as uncertainty rather than strength. Whether that hurts depends on execution.

Decision checklist: how to evaluate this strategy like an analyst or buyer

• □ State the contradiction first: Google competes with Claude but can still profit from it.
• □ Separate economics: Distinguish investment upside from product competition.
• □ Map infrastructure ties: Check the Google Cloud and TPU relationship.
• □ Follow customer demand: Ask which model enterprise buyers actually prefer.
• □ Assess benchmarking carefully: Compare evaluation use with model training claims.
• □ Track the timeline: Separate funding, compute, and distribution milestones.

Use this framework to ask a practical question: is Google optimizing for one winner, or for multiple forms of exposure across the AI stack?

That question is useful beyond Google and Anthropic. It applies to many AI partnerships where the same companies are partly rivals, partly customers, and partly investors.

FAQ

Why is Google investing in Anthropic if it has Gemini?

Because the relationship creates more than one kind of upside. Google can pursue Gemini as its in-house model family while also benefiting from Anthropic through investment exposure, cloud and TPU demand, and enterprise distribution on Google Cloud.

Is Google using Claude to improve Gemini?

Reportedly, Google has used Claude in comparative evaluation workflows. TechCrunch reported that Google contractors were shown Claude responses in internal comparisons, and that Google said DeepMind compares model outputs as part of evaluation but does not train Gemini on Anthropic models.

How much has Google invested in Anthropic?

Public discussion often mixes equity, debt, and infrastructure commitments, so a single number can be misleading. The safest summary from the available materials is that the relationship includes non-voting shares and convertible debt financing, with details evolving over time.

Does Google own part of Anthropic?

Yes, in the sense that UK regulators said Google acquired non-voting shares in Anthropic. But that is not the same as owning or controlling the company. The CMA did not find that Google had acquired material influence over Anthropic.

Is Anthropic a competitor to Google Gemini?

Yes. Claude and Gemini compete for developers, enterprise buyers, and AI workloads. At the same time, Google and Anthropic also have a supplier and distribution relationship through cloud infrastructure and Vertex AI.

What is the relationship between Google Cloud and Anthropic?

Anthropic offers Claude on Google Cloud and through Vertex AI, and it has publicly said it is expanding its use of Google Cloud TPUs and services. That makes Google both a commercial platform partner and an infrastructure provider to Anthropic.

Key takeaways

• Google can profit from Anthropic even if Gemini remains its in-house strategic model family.
• Anthropic is both a competitor and a strategic partner to Google, depending on whether you look at models, cloud, or distribution.
• Cloud and compute economics are central to the relationship; model revenue and infrastructure revenue are not the same thing.
• Enterprise customer demand likely shapes Google’s posture, because buyers often want model choice instead of a single-vendor answer.
• Comparative testing can help Google learn where Gemini is strong or weak without changing the competitive reality.
• The strategy is not irrational once hedging, cloud economics, and enterprise demand are separated into distinct layers.
• The most source-sensitive claims are the exact investment structures, evolving timelines, and reported benchmarking details, which can change as disclosures update.

Read our Claude vs Gemini breakdown for enterprise buyers

References

• https://assets.publishing.service.gov.uk/media/676959bae6ff7c8a1fde9d33/Full_text_decision__.pdf
• https://www.ftc.gov/system/files/ftc_gov/pdf/p246201_aipartnerships6breport_redacted.pdf
• https://claude.com/partners/google-cloud
• https://www.anthropic.com/news/expanding-our-use-of-google-cloud-tpus-and-services
• https://www.googlecloudpresscorner.com/2023-11-08-Google-Announces-Expansion-of-AI-Partnership-with-Anthropic
• https://www.anthropic.com/news/google-vertex-general-availability
• https://docs.anthropic.com/en/api/claude-on-vertex-ai
• https://techcrunch.com/2024/12/24/google-is-using-anthropics-claude-to-improve-its-gemini-ai/