A global shortage of raw computing power is upending Silicon Valley, sparking bitter corporate turf wars, sending executives on worldwide scrambles for hardware, and forging improbable alliances between fierce rivals, according to a report by The Wall Street Journal (WSJ).
The fierce bottleneck has become the tech industry's central battleground, reshaping executive suites and imperiling the prospects of up-and-coming startups that cannot secure enough chips to keep pace.According to a WSJ report, the severity of the compute crunch was laid bare inside Google, which spent $2.7 billion in 2024 to rehire pioneering researcher Noam Shazeer to steer key projects like its flagship Gemini models.
Despite the enormous price tag, Google failed to guarantee Shazeer the computing capacity he sought. After internal reallocations cut his team’s resources in favor of another initiative, Shazeer departed for rival OpenAI, highlighting an internal environment where researchers frequently quarrel over processing power.
The Journal reported that Google co-founder Sergey Brin has at times overridden formal allocation procedures to divert compute to his preferred projects, while DeepMind co-founder Demis Hassabis has expressed frustration over a lack of resources dedicated to fundamental, long-term research.
According to an earlier report in CNBC, "One of the biggest points of friction inside Google is compute." The report added, "Google is investing more than almost any company in the world in data centers, chips and related infrastructure. But capacity remains scarce. Every TPU assigned to training a model, serving a Google product, or fulfilling a contract with a cloud customer reflects a choice among competing priorities.
" Claiming sources it further said that frustrations over access to compute can be especially acute when Google announces large infrastructure commitments to competing labs like Anthropic, whose models compete directly with Gemini.
Elon Musk publicly reverses his stand on Anthropic, days after slamming it
The scramble for capacity has driven competitors into surprising pacts, most notably between Anthropic and Elon Musk's xAI. Facing severe infrastructure strain and outages triggered by the runaway popularity of its Claude Code assistant, Anthropic dispatched co-founder Tom Brown to xAI’s offices in Hawthorne, California, in late March to negotiate renting compute directly from Musk’s venture.
The talks paved the way for Anthropic to lease over 300 megawatts of computing power through an agreement connected to Musk’s company, a deal valued at up to $45 billion.
The alliance bridged an ideological divide between Musk and Anthropic chief executive Dario Amodei. Just days before the partnership, Elon Musk, in a post on X, wrote, ""Your AI hates Whites & Asians, especially Chinese, heterosexuals and men. This is misanthropic and evil," Musk wrote.
"Fix it."
Anthropic's Dario Amodei knocks at rival Meta's door
Amodei also reportedly approached Meta’s chief AI officer Alexandr Wang in search of excess capacity. According to the report, "Anthropic CEO Dario Amodei also spoke on the phone with Meta chief AI officer Alexandr Wang earlier this year, attempting to source more computing capacity, according to a person familiar with the matter. Meta debated whether to give chips to Anthropic and ultimately decided not to for now, people familiar with the matter said.
"Across the wider tech sector, the hunt for computational power resembles a frantic gold rush. Intermediaries like San Francisco Compute are scouring unconventional sites worldwide to locate usable clusters, finding servers running in places as unusual as a converted chicken coop in rural America or navigating monsoon floods in Manila to source hardware in Taipei. Simultaneously, the market has swung aggressively in favor of suppliers: the hourly cost to rent Nvidia H100 chips on one-year contracts jumped 60 percent over the past year, while cloud providers now routinely demand multiyear leases and massive upfront down payments.