No room for pessimists in the AI buildout

Welcome to Runtime! Today: AI companies are making huge spending commitments that require an enormous leap of faith in their revenue-generating abilities, two cybersecurity professionals are indicted in a series of ransomware attacks, and the latest funding rounds in enterprise tech.

(Please forward this email to a friend or colleague! And if it was forwarded to you, sign up here to get Runtime each week.)

Long way to go and a short time to get there

Some truly astonishing numbers are being thrown around as we come to the end of the beginning of the generative AI boom. If cloud and AI companies somehow manage to follow through on the enormous deals they've made for computing infrastructure services over the past few weeks, they'll have pulled off one of the most incredible feats in tech history.

AWS finally welcomed OpenAI into its arms Monday with a $38 billion deal for its computing services. After Microsoft and OpenAI formally modified their agreement to allow OpenAI to find computing capacity from other companies, it felt like only a matter of time before it tapped AWS for capacity after signing a deal with Google Cloud earlier this year.

• The companies said the deal will run for seven years but OpenAI will start using AWS's arsenal of Nvidia GPUs immediately, "with all capacity targeted to be deployed before the end of 2026, and the ability to expand further into 2027 and beyond."
• Theory Ventures' Tomasz Tunguz did the math, and OpenAI has now committed to spend $1.15 trillion on computing services over the next decade.
• "We’re in the middle of probably the largest technology buildout that we’ve ever seen," Lambda CEO Stephen Balaban told CNBC Monday.

But there's one problem with this buildout: An awful lot of these spending commitments are backloaded for 2027 or beyond, and depend on companies like OpenAI and Anthropic to find ways to grow revenue at an unfathomable speed.

• Tunguz estimated that OpenAI would have to find a way to generate $111 billion in revenue during 2027 and $295 billion in 2028 to support its spending commitments over those years.
• For comparison, Amazon CEO Andy Jassy said last week that AWS — a 15-year-old business that counts an enormous number of the world's enterprise companies as a customer —  is on pace to do $132 billion in revenue this year, and Microsoft just recorded $281.7 billion in revenue during its most recent fiscal year, which ended in June.
• Meanwhile, OpenAI expects to record about $13 billion in revenue this year.

Hitting those targets will require basically the entire enterprise computing world to decide to devote a big chunk of their IT budgets to accessing large-language models through APIs, or hiring OpenAI and Anthropic to build those systems for them, in the next two years. Gartner expects those IT budgets to grow by just 9.8% next year, which makes it a little hard to understand where the money will come from.

• While there's no question demand for generative AI services has grown over the last three years, most companies are still trying to figure out how and where those expensive services fit into their business models.
• Brad Gerstner of Altimeter Capital asked OpenAI CEO Sam Altman about its spending commitments and revenue projections on a podcast over the weekend, and Altman's annoyed response to having his plans questioned was quite interesting.
• "We are taking a forward bet that [revenue is] going to continue to grow and that not only will ChatGPT keep growing, but we will be able to become one of the important AI clouds, that our consumer device business will be a significant and important thing, that AI can automate science [and] will create huge value," he said, according to Fortune.

So, to recap: All OpenAI has to do to generate enough revenue to cover its spending commitments is to become one of the world's most important enterprise, consumer, and scientific companies in the next couple of years.

• A lot can happen in a couple of years, but that would be quite an achievement.

The call is coming from inside the house

Cybersecurity executives worry constantly about insider threats to their data or operations, but according to a new indictment released this week, now they'll have to wonder whether the people they've hired to protect those assets are working both sides of the street. Federal prosecutors Monday accused three cybersecurity professionals of launching ransomware attacks against several businesses while helping other companies defend themselves against ransomware attacks.

Two of the men worked for Digital Mint, which "[provides] swift, compliant solutions for a safer digital future," and the other worked for Sygnia Cybersecurity Services, which promises customers that "we understand how malicious actors think and behave." The complaint does not suggest that the companies had any idea what their employees — who have since been fired — were doing, but it's not a great look.

If there's one silver lining in this story, it's that businesses are getting the message about refusing to pay the ransom when confronted with these kinds of attacks. The defendants targeted several businesses but only one paid $1.2 million in ransom, according to The Chicago Sun-Times.

Enterprise funding

Beacon Software raised $250 million in Series B funding for its private-equity-like approach to acquiring small companies and rebuilding them around AI.

Applied Compute launched with $80 million in funding and three former OpenAI engineers to help companies build AI agents tailored around their specific business model and data.

Daylight landed $33 million in Series A funding to build out a new type of  managed-detection-and-response security platform around agentic AI.

CoreStory raised $32 million in Series A funding to help companies refactor old code with AI coding agents.

Teleskope landed $25 million in Series A funding for its AI security platform, which automatically detects misconfigured infrastructure.

Reflectiz scored $22 million in Series B funding to build out a security platform that helps companies understand how their web sites are exposed to threats.

The Runtime roundup

AMD's data-center revenue grew 22% and beat Wall Street estimates, but the after-hours crowd didn't like its growth forecast for the fourth quarter.

AWS announced plans to deploy a fiber-optic cable between County Cork in Ireland and Maryland, and it is expected to light up in 2028.

Google plans to launch a handful of its TPU AI chips into space by 2027 in hopes of figuring out a way to take advantage of unlimited solar power, according to Semafor.

IBM announced plans to cut thousands of jobs this quarter, impacting what it called "a low single-digital percentage of our global workforce," which has more than 270,000 people.

Thanks for reading — see you Thursday!