OpenAI courts Amazon as Microsoft tension grows

OpenAI says its new Amazon Web Services deal is already pulling in a wave of enterprise demand, and the timing is hard to miss. In a memo to staff, revenue chief Denise Dresser said Microsoft has “limited our ability” to meet customers where they are, especially inside Amazon Bedrock.

This is more than a partner update. It is a window into how OpenAI is trying to sell into corporate IT while also loosening its dependence on Microsoft, its biggest backer and longest-running cloud ally.

And the numbers are large enough to matter. Amazon said in February that it plans to invest up to $50 billion in OpenAI as part of the partnership, while Microsoft has already poured more than $13 billion into the company since 2019. OpenAI’s enterprise business now makes up 40% of revenue, according to Dresser, and she says it is on track to match consumer revenue by year-end.

Why Amazon matters so much to OpenAI

OpenAI’s enterprise push lives or dies on distribution. Corporate buyers do not want to rebuild their stack around a single vendor, and many already buy their AI tools through AWS procurement, security, and billing workflows. That is why Bedrock matters: it lets customers access multiple foundation models through one AWS interface, including OpenAI models.

Dresser’s memo makes the strategy plain. She said the Microsoft relationship has been “foundational” but also restrictive, because plenty of enterprises buy through AWS first. If OpenAI wants to win larger contracts, it has to show up where those buyers already spend money.

The immediate signal is demand. Dresser wrote that inbound interest since the Amazon partnership was announced has been “frankly staggering.” That is the kind of language companies use when sales teams are seeing a real pipeline shift, not just a press-release bump.

• Amazon plans to invest up to $50 billion in OpenAI.
• Microsoft has invested more than $13 billion in OpenAI since 2019.
• OpenAI says enterprise accounts for 40% of revenue.
• Dresser says enterprise revenue is on track to reach consumer parity by the end of 2026.

Microsoft is still central, but the friction is real

OpenAI and Microsoft still describe their relationship as strategic, and that is true in the narrow sense that Microsoft remains deeply embedded in OpenAI’s compute and distribution story. But the partnership has clearly become more complicated as both companies move deeper into enterprise AI.

Microsoft has also started treating OpenAI more like a peer than a protected favorite. In mid-2024, Microsoft added OpenAI to its annual list of competitors, alongside Amazon, Apple, Google, and Meta. That is a subtle but meaningful signal: the two companies are partners, but they are also increasingly rivals in AI software, cloud delivery, and developer tooling.

“Our Microsoft partnership has been foundational to our success. But it has also limited our ability to meet enterprises where they are — for many that’s Bedrock.” — Denise Dresser, OpenAI revenue chief, in an internal memo viewed by CNBC

That quote tells you exactly where the tension sits. OpenAI wants the scale and credibility that Microsoft brings, but it also wants the freedom to sell through other channels without asking permission from the same company that helped fund its rise.

Enterprise AI is turning into a cloud fight

The broader market context matters here. Enterprise AI is where the money is, because companies are spending heavily on copilots, search, coding assistants, and internal workflow tools. That spending is also pressuring public software firms, many of which are now being judged against AI-native competitors instead of legacy SaaS peers.

That is why OpenAI’s battle with Anthropic and Google is really a cloud-channel battle too. Anthropic has become a favorite inside many enterprise teams, in part because Claude is widely available through AWS Bedrock and Google Cloud. Google’s Cloud and Vertex AI are also pushing hard to keep model choice inside their own ecosystems.

The recent numbers show how fast the competition is moving:

• OpenAI was valued at more than $850 billion in its latest fundraising round in late March.
• Anthropic was valued at $380 billion in the round that followed.
• Anthropic said earlier this month that its run-rate revenue has passed $30 billion.
• OpenAI says Anthropic’s revenue figure is inflated by roughly $8 billion because of accounting treatment tied to partner rev share.

Those are not small gaps. They show a market where valuation, distribution, and cloud access are all tied together. Whoever controls the enterprise buyer relationship gets a lot more than model usage; it gets sticky recurring revenue and a better shot at becoming the default AI layer inside the company.

What this says about OpenAI’s next move

Denise Dresser joined OpenAI in December after running Slack and spending years in senior roles at Salesforce, which makes her the kind of operator who thinks in pipeline, conversion, and enterprise packaging. Her memo reads like a sales leader trying to reframe the company’s identity from consumer breakout to serious enterprise vendor.

OpenAI is also showing that it wants optionality. It has already turned to CoreWeave, Google Cloud, and Oracle Cloud for more capacity. That matters because compute access is now a sales strategy, not just an infrastructure detail.

One more piece is worth watching: Anthropic’s own expansion. The company recently announced a deal with Google and Broadcom for multiple gigawatts of compute. That kind of capacity commitment can change how fast a model company can ship, serve customers, and defend its margins.

My read: OpenAI is trying to become the default AI vendor for companies that do not want to marry one cloud provider. If the Amazon channel keeps producing demand, expect more enterprise packaging, more multi-cloud selling, and more public friction with Microsoft. The real question is whether OpenAI can keep that balance while still relying on Microsoft for core infrastructure and distribution. If it can, its enterprise business may catch consumer revenue sooner than the company is letting on.