Buying data centers is easier than building them

Welcome to Runtime! We're back from vacation! Today: Why CoreWeave just shelled out $9 billion in stock for Core Scientific, Ingram Micro begins to recover from a holiday weekend ransomware attack, and the latest funding rounds in enterprise tech.

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

Counting our real estate

Just a few years ago it seemed impossible to imagine that a startup could hope to dent the reach of the Big Three cloud infrastructure providers given the enormous cost of building a global data-center footprint from scratch. While the generative AI frenzy opened a door for "neocloud" GPU-focused challengers like CoreWeave, the cost of playing this game goes up every year.

CoreWeave announced Monday that it plans to acquire Core Scientific for $9 billion in stock, taking advantage of a 277% increase in the price of that stock since its IPO in March. The deal would allow CoreWeave to take direct control over infrastructure it is already renting from Core Scientific and expand its future capacity without having to stand up new data-center buildings on its own, executives said on a conference call Monday.

• Core Scientific, which is primarily known for supporting Bitcoin miners, owns 1.3 gigawatts of data-center capacity across the U.S., 840 megawatts of which is already under contract to CoreWeave.
• As of the end of last year according to its S-1 registration statement, CoreWeave directly controlled just 360 megawatts of data-center capacity, renting the rest from other providers.
• The deal will allow CoreWeave to stop paying rent for Core Scientific's servers and convert an additional 500 megawatts of capacity allocated for Bitcoin mining to AI infrastructure, according to executives.
• "Owning Core Scientific's high performance data center infrastructure enables us to significantly enhance operating efficiencies and de-risk our future expansion," CEO Michael Intrator said on the conference call.

By and large, the Big Three cloud hyperscalers funded the construction of their massive data-center portfolios with revenue from their main businesses, namely online retail, enterprise software, and search advertising. CoreWeave, on the other hand, funded its capacity expansion through what it has described as its "unique financing capabilities," such as taking out more than $14 billion in debt tied to the value of the Nvidia GPUs it owns.

• At some point it will need to service that debt, which the Financial Times reported earlier this year could require $7.5 billion in payments by the end of 2026.
• Intrator told financial analysts that acquiring Core Scientific would "drive significant cost savings by streamlining core business operations, eliminating third-party lease obligations, and unlocking greater control over fixed infrastructure costs."
• The deal would allow Coreweave to shed $10 billion in costs that were already allocated toward future data-center lease agreements and find as much as $500 million in annual savings by 2027, according to chief financial officer Nitin Agrawal.

Core Scientific will also give CoreWeave the additional capacity it needs to fulfill nearly $16 billion in orders from OpenAI over the next several years without having to raise new debt or find suitable locations for brand-new datacenters, which is becoming increasingly difficult. However, CoreWeave's long-term chances of enduring as an AI hyperscaler will depend less on financial wizardry and more on its ability to convince average enterprises to bet on its platform.

• CoreWeave's IPO revealed how much it depended on Microsoft for revenue, and while recent deals with OpenAI and Google help diversify its customer base, all three of those companies will continue to build their own AI infrastructure.
• CoreWeave is betting that it can attract cost-conscious customers by investing in a software stack that allows customers to utilize those expensive GPUs more efficiently than the Big Three, Chen Goldberg, senior vice president of engineering, told Runtime in an interview earlier this year.
• Core Scientific might help CoreWeave solve some of its financial challenges, but only diversified and sustainable revenue growth will make it a true challenger to the major hyperscalers.
• And that growth will depend on how demand for enterprise AI models, applications, and agents is playing out, which we'll learn more about over the next several weeks as enterprise tech puts the finishing touches on its second-quarter earnings reports.

Wave that flag

After a parade of high-profile ransomware attacks disrupted several different companies last year, 2025 had been relatively quiet on the ransomware front until just before the holiday weekend last week. Ingram Micro confirmed Sunday that a prolonged systems outage that began Thursday morning was related to a ransomware attack, and as of publishing time Tuesday afternoon the recovery operation was still in progress.

Bleeping Computer reported Saturday that the SafePay ransomware gang appeared to be behind the attack, which brought the operations of one of the biggest IT distributors on the planet to a halt. On Monday Ingram Micro said it was able to take some orders over the phone from customers in several countries, but the U.S. was not listed as one of those countries as of a Tuesday update that noted "the Company proactively chose to take certain systems offline as part of our mitigation efforts, processes, and protocols associated with this cybersecurity incident."

That's a step many other ransomware victims have chosen to take after detecting an attack, such as MGM Resorts, but it's not clear why a company so tightly integrated with the global IT community didn't have a faster way to get back up and running. And obviously, for a substantial portion of its customer base Friday was just another work day.

Enterprise funding

Campfire raised $35 million in Series A funding for its ERP software, which was designed to take on legacy services from companies like SAP and Oracle.

Wonderful scored $34 million in seed funding to build AI customer-support tools for companies in markets where English is not the primary language.

Arago landed $26 million in seed funding for its AI chip design, which uses optical technology to reduce the amount of energy consumed by an AI workload.

Emerald AI raised $24 million in seed funding and unveiled its energy-management software, which orchestrates AI workloads in data centers to reduce energy use.

Sundial scored $16 million in Series A funding for its business analytics platform.

Pokee AI landed $12 million in seed funding as it develops an AI-agent building platform to help automate business workflows.

The Runtime roundup

Oracle extended a 75% discount on its core software products to the federal government, according to The Wall Street Journal, which is surely not the kind of gesture for which it will expect something in return.

Cursor apologized for botching the rollout of new pricing plans for its AI coding tool, which hit a number of customers with unexpected charges in June.

Thanks for reading — see you Thursday!