Amazon Web Services is placing enormous bets on artificial intelligence — and it doesn’t much care if those bets compete with each other.
Matt Garman, the CEO of AWS, recently offered his clearest explanation yet for why the cloud giant is simultaneously investing billions of dollars in Anthropic while also courting OpenAI as a customer and partner. His argument boils down to something deceptively simple: AWS is a platform, and platforms don’t pick favorites. They sell to everyone.
The logic sounds tidy. Whether it holds up under the extraordinary pressures of the AI arms race is another question entirely.
As TechCrunch reported, Garman addressed the apparent tension head-on during a recent press event, framing Amazon’s strategy as consistent with how AWS has always operated. The company has long hosted competing software vendors on its cloud infrastructure. It runs Oracle databases alongside its own Aurora. It supports Microsoft SQL Server even as it pitches homegrown alternatives. From Garman’s perspective, adding rival AI model makers to the mix is just the next chapter of the same playbook.
But the scale of the financial commitments here dwarfs anything AWS has done before with third-party software companies. Amazon has invested up to $4 billion in Anthropic, the Claude developer founded by former OpenAI researchers Dario and Daniela Amodei. That investment came with a deal making AWS Anthropic’s primary cloud provider and giving Amazon’s custom chips a prominent role in training Anthropic’s models. It was, by any measure, a deep strategic alignment — the kind of partnership where the two companies’ futures become partially intertwined.
Then came OpenAI.
Reports earlier this year confirmed that OpenAI, long synonymous with Microsoft Azure, had begun running some workloads on AWS infrastructure. The move surprised industry observers who had assumed OpenAI’s relationship with Microsoft — which has poured more than $13 billion into the Sam Altman-led company — was essentially exclusive. It wasn’t. And AWS was happy to be the beneficiary of that loosening arrangement.
Garman’s framing treats this as unremarkable. “We want to be the best place for any AI company to build,” he said, according to TechCrunch’s account of his remarks. The implication is that AWS’s role as neutral infrastructure provider supersedes any competitive awkwardness created by its Anthropic investment. Customers choose the models they want. AWS just provides the compute.
That framing has limits, though, and Garman seems to know it. Anthropic isn’t just another tenant on AWS servers. Amazon holds a significant equity stake. Anthropic’s models are featured prominently in Amazon Bedrock, the company’s managed AI service that lets enterprise customers access foundation models through a single API. Anthropic’s Claude has become, in many ways, the flagship model of Amazon’s AI strategy — the answer Amazon gives when customers ask which large language model they should use if they don’t want to depend on Google or Microsoft.
So when OpenAI shows up on the same infrastructure, it creates a dynamic that’s more complicated than Garman’s platform-neutrality argument suggests. Anthropic’s leadership has to wonder, at least privately, whether their most important financial backer is diluting the partnership’s value by rolling out the red carpet for their biggest competitor. OpenAI’s leadership, meanwhile, has to weigh whether running on AWS means indirectly enriching a company that’s bankrolling their rival.
Neither side has publicly complained. That silence speaks to the pragmatism driving all parties. In AI right now, compute is king. Access to enough GPUs — and increasingly, to custom accelerators like Amazon’s Trainium chips — matters more than competitive purity. If OpenAI can get better pricing or availability on AWS for certain workloads, it would be irrational to refuse on principle. And if Anthropic’s models get wider distribution because AWS is seen as a credible multi-model platform, the exposure is worth tolerating a neighbor they’d rather not have.
The financial math reinforces this pragmatism. AWS generated over $100 billion in annual revenue in its most recent fiscal year, and AI workloads are the fastest-growing segment of that business. Every major AI lab running on AWS translates directly into higher utilization of Amazon’s data centers, more demand for its networking infrastructure, and more leverage in negotiating chip supply deals with Nvidia and others. Garman doesn’t need Anthropic or OpenAI to win the model wars. He needs them both to keep buying compute.
This is the core of the strategy, and it mirrors what Nvidia has done in hardware. Jensen Huang’s company sells GPUs to every AI lab, every hyperscaler, every startup with enough funding to place an order. Nvidia doesn’t care who wins. It profits from the competition itself. Garman appears to want AWS to occupy an analogous position in cloud infrastructure — the arms dealer who sells to all sides.
There’s a crucial difference, though. Nvidia doesn’t own equity in any of its major customers. Amazon does. That $4 billion Anthropic stake means Amazon’s financial interests aren’t perfectly neutral. If Anthropic becomes the dominant AI model provider, Amazon’s investment appreciates enormously. If OpenAI wins instead, Amazon still collects cloud revenue, but the Anthropic investment loses relative value. The incentives aren’t symmetrical, no matter how Garman frames them.
Wall Street has largely shrugged at the contradiction. Amazon’s stock has performed well as investors have bought into the narrative that AI spending will benefit cloud providers regardless of which model companies emerge as long-term winners. Analysts at firms including Goldman Sachs and Morgan Stanley have pointed to AWS’s AI revenue growth as validation of the multi-model approach. The market, for now, is willing to accept that platform economics can coexist with strategic investments.
But the AI industry is changing fast, and the relationships that seem manageable today could become untenable tomorrow. Consider a scenario where Anthropic and OpenAI begin competing aggressively for the same enterprise customers on AWS’s Bedrock platform. Would AWS’s recommendation algorithms and sales teams remain truly neutral? Would Anthropic’s models get preferential placement, given the equity stake? Or would AWS bend over backward to appear fair, potentially disadvantaging its own investment?
These aren’t hypothetical concerns. Google Cloud faces similar questions with its own AI partnerships, and Microsoft has already navigated tensions between its OpenAI alliance and its desire to offer competing models from Mistral and Meta on Azure. The multi-model strategy is becoming industry standard among hyperscalers, but Amazon’s Anthropic investment makes its version of that strategy uniquely fraught.
Garman’s public confidence suggests he believes the market is big enough to absorb these tensions. And he may be right — for now. Enterprise AI adoption is still in early innings, with most large companies experimenting rather than committing to a single model provider. In that environment, a platform offering multiple options has a natural advantage. IT buyers like choice. They like knowing they aren’t locked in.
The real test will come when the market matures and enterprises start consolidating around fewer models. At that point, AWS will have to decide whether it’s truly a neutral platform or a company with a preferred AI partner. The billions invested in Anthropic suggest the answer is already determined, whatever the official messaging says.
For Garman, the immediate priority is growth. AWS’s AI-related revenue has been accelerating, and the company has announced plans to spend approximately $100 billion on capital expenditures in the current fiscal year, much of it on data centers designed for AI workloads. That spending spree reflects a conviction that demand for AI compute will continue to outstrip supply for years to come. In a supply-constrained market, having both Anthropic and OpenAI as customers is simply good business.
And Amazon isn’t alone in embracing this kind of strategic ambiguity. The entire cloud industry is moving toward a model where infrastructure providers maintain relationships with competing AI companies while simultaneously building their own AI capabilities. It’s messy. It’s contradictory. And for the moment, it’s enormously profitable.
The question that hangs over all of it — the one Garman didn’t directly answer — is what happens when the interests genuinely collide. When Anthropic asks for exclusivity. When OpenAI demands guarantees that its proprietary training data won’t be accessible to a competitor backed by the same cloud provider. When enterprise customers start asking pointed questions about data isolation and competitive firewalls.
Those conversations are coming. They may already be happening behind closed doors. And when they surface publicly, Garman’s tidy narrative about platform neutrality will face its most serious test.
Until then, AWS will keep cashing checks from both sides. Billions flowing in from Anthropic’s cloud consumption. Billions more from OpenAI’s expanding footprint. And a $4 billion equity bet that, for all the talk of neutrality, reveals exactly where Amazon thinks the future of AI is headed.
WebProNews is an iEntry Publication