AWS and OpenAI come full circle

Welcome to Runtime! Today: AWS finally ties the knot with OpenAI after a long courtship, GitHub remains trapped in a cycle of downtime, and the latest funding rounds in enterprise tech.

Please forward this email to a friend or colleague! If it was forwarded to you, sign up here to get Runtime each week, and if you value independent enterprise tech journalism, click the button below and become a Runtime supporter today.

Agent of change

SAN FRANCISCO — More than three years after Microsoft and OpenAI caught AWS off-guard with the launch of ChatGPT and surged to an early lead in enterprise AI thanks to their exclusive partnership, the cloud leader has leveled the playing field. AWS announced Tuesday that OpenAI's models will soon be available through AWS's Bedrock service without restrictions, a big boost for its AI strategy.

"This is a huge partnership, and it's one of the things that we are quite excited about," AWS CEO Matt Garman said Tuesday at an event in downtown San Francisco. AWS will now host all of OpenAI's models  — not just the open-source ones — as well as its Codex coding agent on Amazon Bedrock, a deal that was only possible after Microsoft and OpenAI agreed on Monday to amend the agreement that gave Microsoft the exclusive right to host OpenAI's models.

• Denise Dresser, OpenAI's relatively new chief revenue officer, signaled this expanded partnership was coming in a memo to employees earlier this month when she wrote that OpenAI's then-current agreement with Microsoft "limited our ability to meet enterprises where they are – for many that’s Bedrock."
• With the new agreement, "this is as easy as a click of a button now to build with access to models, to the API and Codex, and I think this just unlocks speed and scale to really operate," Dresser said at the event Tuesday.
• The models and Codex are only available in "limited preview" at the moment, which likely means access is being doled out to the big accounts.

AWS and OpenAI had already started working together more closely back in February with the introduction of the Stateful Runtime Environment, which is now called Amazon Bedrock Managed Agents. It appeared at the time to be a clever way to circumvent the fact that Microsoft was the exclusive provider of stateless API calls to OpenAI's models under their existing agreement, but AWS principal engineer Anthony Liguori insisted Tuesday that the stateful approach really will be the best architecture for enterprise agents.

• "What I'm really focused on and what really excites me, just earnestly excites me, is how do we go from right now, 'AI is transforming software engineering,' to having AI transform the rest of the industry," Liguori said in an interview with Runtime.
• He thinks that road map requires "stateful" AI agents, which can be run in the cloud and offer more security than "stateless" agents, where context is stored locally and has to be delivered to the AI model with every query.
• This is especially important for customers in regulated industries or large enterprises with lots of different role-based access controls, and the stateful approach also makes it easier to chain agents together to execute multistep tasks, he said.
• However, Liguori acknowledged that AWS doesn't offer a stateful product that works with models offered by Anthropic, which until Tuesday was AWS's AI model bestie.

No matter the state of play, the expanded deal will boost enterprise business for both companies, and expect Amazon CEO Andy Jassy to take quite the victory lap Wednesday during Amazon's first-quarter earnings call. "Already hearing this deal with AWS is negative to Azure," Creative Strategies' Ben Bajarin posted on X after the press conference, although that might be a little premature.

• But Tuesday's announcement showcased just how much has changed in enterprise AI over the last three years, back when AWS was constantly on the defensive in a market that seemed to have declared Microsoft a winner thanks to that special relationship with OpenAI.
• "This is what our customers have asked for for a really long time," Garman said during the event. "Their production applications run in AWS. Their data is AWS. They trust the security of AWS, and we forced them, for the last couple of years, to have to get the great OpenAI models, to go to other places." 

Null request

There is no question that demand for enterprise AI and software development services has exploded over the last six months, worsening the reliability problems that plagued both Anthropic and OpenAI last year. Those companies, however, are attempting to serve a very new and complicated type of workload across multiple types of hardware, whereas GitHub appears to have completely failed to anticipate how its current customers would respond to the AI coding agent boom it helped introduce.

GitHub was forced to publish another blog post Tuesday explaining two severe incidents from the past week that only added to the growing frustration software developers are having with the nearly ubiquitous coding hub. Last Thursday developers struggled to deal with a merging issue that in some cases overwrote previous changes to their repositories, and on Monday GitHub's search function went down, preventing developers from finding outstanding pull requests and issues.

"Both of those incidents are not acceptable, and we are sorry for the impact they had on you," the company said in the post. GitHub pledged to increase computing resources in order to keep up with what it said was a 30x increase in demand for its services, and it sounds like it has resorted to paying one of Microsoft's competitors to stay afloat: "While we were already in progress of migrating out of our smaller custom data centers into public cloud, we started working on path to multi cloud."

Enterprise funding

VAST Data raised $1 billion in Series F funding, which values the AI infrastructure company at $30 billion.

Quantum Art landed $140 million in an extension to a previous Series A round for its quantum computing strategy, which focuses on scalability.

Omni scored $120 million in Series C funding for its data-management platform, which values the company at $1.5 billion.

Cloudsmith raised $72 million in Series C funding for its artifact management platform, which helps companies secure their software supply chains.

Orkes landed $60 million in Series B funding for its workflow orchestration platform, which was developed by former Netflix engineers responsible for its microservices orchestration platform.

exe.dev scored $35 million in Series A funding to "build a new generation of cloud infrastructure" designed for the AI coding agent era.

The Runtime roundup

AWS also launched two new applications Tuesday during the event, an OpenClaw/Claude Cowork-like desktop app called Amazon Quick and an expansion of its Amazon Connect app into several other enterprise categories, as it continues to balance its traditional role providing building blocks for enterprise apps with the desire from some customers to buy a complete product.

F5 beat Wall Street's expectations and raised revenue guidance for the full year, citing an increase in hybrid cloud strategies, security threats, and, of course, AI applications.

Thanks for reading  — see you Thursday!