What is Microsoft going to do with OpenAI?

Welcome to Runtime! Today: OpenAI's attempts to alter its sweeping deal with Microsoft will force some interesting decisions in Redmond, Google Cloud offers more details about last Thursday's outage, and the latest funding rounds in enterprise tech.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Not Azure thing

Microsoft and OpenAI's landmark 2019 deal allowed Microsoft to vault ahead of its enterprise competitors following the launch of ChatGPT in 2022, which created enormous demand for OpenAI's large-language models on Microsoft's cloud infrastructure service and introduced OpenAI to a generation of enterprise tech leaders. But a lot has changed since that deal went down, and both companies are now preparing for a future as competitors.

Evidence of tension between the two companies has built ever since OpenAI CEO Sam Altman was fired and rehired during a bizarre week in 2023, which appeared to catch Microsoft CEO Satya Nadella off guard. On Monday, separate reports from The Wall Street Journal and The Information outlined how the two companies are haggling over the terms of OpenAI's conversion to a traditional for-profit startup, which it must do by the end of the year to obtain $40 billion in funding from Softbank.

• Under the original deal, Microsoft has the exclusive rights to offer OpenAI's models on Azure, but OpenAI wants to significantly expand the number of customers it can reach by working with AWS and Google Cloud.
• In exchange, Microsoft wants a larger stake in the for-profit company than OpenAI is willing to offer, according to the reports.
• Microsoft also wants to extend the length of time it has access to OpenAI's intellectual property, which under the current deal runs through 2030, according to The Information.

The IP question appears to be a key part of this dispute, and helps explain why OpenAI's pending acquisition of AI coding startup Windsurf, first reported over a month ago, has yet to close. According to both reports, OpenAI wants to prevent Microsoft from having access to Windsurf's IP.

• Windsurf is a fast-growing startup that competes in the AI coding editor market with Cursor, which just raised $900 million in funding valuing the company at a shade under $10 billion.
• However, both Windsurf and Cursor are forks of Microsoft's open-source Visual Studio Code, which means both editors rely on VS Code for their internal plumbing and compete through the user interface and features they layer on top of that code.
• If Microsoft had access to Windsurf's IP, it could build a free or low-cost version of VS Code that could make it hard to justify paying for Windsurf, which costs $60 per user per month for an enterprise version.
• Or it could assign that IP to GitHub Copilot, which also competes with Windsurf.

One major factor that has changed since 2022 is access to world-class LLMs from several companies, and the value in enterprise AI has shifted to the vendors that are building the tools around those models that help enterprises use them in their applications. Companies are realizing that they can use different models for different types of applications, and products such as Windsurf allow customers to easily switch between models from OpenAI, Anthropic, Google, and others.

• OpenAI's models are still among the best at what they do and AWS and Google would absolutely offer them to their customers if they could, but exclusive access to OpenAI's models is less of a competitive advantage among enterprise buyers in 2025 than it was three years ago.
• Microsoft has been growing closer to Anthropic since last year, and last month's announcement that Anthropic's newest Claude Sonnet models would be an option in a new coding agent feature might have been an acknowledgement that its days of OpenAI exclusivity are numbered.
• Still, it's clear that Microsoft values its access to OpenAI's IP to help inform its product-development strategy over the rest of the decade, and extending that access into the next decade is worth something.
• Given that OpenAI's valuation is based on a magical revenue projection that will require it to grow more than 800% over the next five years, it might make sense to lock in IP access in exchange for a smaller stake in a potentially doomed venture.

Wave that flag

Cloud infrastructure service providers have made great strides toward reducing the frequency and impact of cloud outages over the last decade, and Google has long been at the forefront of those advances. But Google Cloud acknowledged late Friday afternoon that the widespread outage endured by its customers last Thursday could have been easily prevented.

On May 29th, Google Cloud updated the system that checks how often its customers are accessing its APIs, which is how cloud providers bill customers for usage. However, when it updated that system, the new code "did not have appropriate error handling nor was it feature flag protected," which would have allowed engineers to quickly turn off anything that was causing problems without having to redeploy code.

And so when another change was made Thursday morning to Google's Spanner database that "contained unintended blank fields," it was automatically sent to Google servers around the world and created what's called a "null pointer" problem, which happens when software expects to find something in a field that's not there and freaks out. If Google had followed its usual policies and added feature flags to the change rolled out in May, "the issue would have been caught in staging."

Enterprise funding

Ramp raised $200 million in Series E funding, valuing the financial operations company at $16 billion.

Meter landed $170 million in Series C funding for its networking infrastructure technology, which uses a combination of hardware and software to help companies build networks on-premises.

Coralogix scored $115 million in Series E funding and launched a new AI agent called Olly, which can't do a kickflip but allows companies to use natural-language commands with the company's observability tools.

Browserbase raised $40 million in Series B funding and introduced a product called Director.ai, which makes it easier to build AI agents that use its headless browser to navigate the web.

Fleet landed $27 million in Series B funding for its open device management software, which helps companies manage all the end-user computers on their networks.

Sedai scored $20 million in Series B funding to help companies manage their cloud infrastructure using autonomous AI agents.

The Runtime roundup

Belts are getting tighter at Salesforce, which raised prices across the board by around 6% and kicked the Kubernetes community off its free Slack enterprise account.

AWS announced Tuesday that 100% of users who have root access to accounts are now using multifactor authentication, which is a big number and quickly becoming the standard for enterprise tech after last summer's Snowflake debacle.

Thanks for reading — see you Thursday!