In an unexpected organizational paradox, researchers at one of the world’s largest tech companies are suffering from a shortage of computing resources that their company provides in abundance to external competitors.
- Massive infrastructure and internal paradox
- Impact of massive cloud contracts and competitors
- Management acknowledgment and global supply chain crisis
- Experts leave to establish new companies and labs
- Expert assessment of the landscape and resource allocation
- Frequently asked questions
Massive infrastructure and internal paradox
Over past years, tech giant Google has successfully built and developed the strongest and most advanced infrastructure in the technology sector to meet the demands of the digital revolution. This sophisticated arsenal includes custom, high-performance computing chips designed in-house, a booming cloud sector generating substantial profits, and multi-billion-dollar infrastructure supply deals with prominent competitors in the tech market. However, this dazzling commercial and strategic success has created an unexpected internal paradox and an unprecedented crisis: the company’s creative internal researchers, who are relied upon to drive future innovation, cannot secure enough time and resources to run their complex experiments on their own company’s hardware. This has significantly impacted their advanced research path and hindered their accelerated scientific ambitions.
Impact of massive cloud contracts and competitors
Extensive news reports published by Bloomberg on Monday revealed that researchers and scientists in the esteemed Google DeepMind division are competing fiercely and with extreme difficulty to gain stable access to precious computing resources—the very same resources that the company sells and leases to external clients under long-term contracts. The problem and structural reason behind this crisis lie in strict legal and contractual commitments. The company has committed to providing a massive infrastructure totaling $40 billion to Anthropic under a historic, mega deal. This agreement includes supplying massive processing unit capabilities of 5 gigawatts over five consecutive years, with a commitment to guarantee access to one million highly advanced seventh-generation chips. A similar strategic arrangement also serves Meta’s requirements. These large institutional commitments tie up massive computing capacities that become unavailable to internal model development teams until they wait in long lines that disrupt research progress.
Management acknowledgment and global supply chain crisis
In an attempt to manage this complex situation, Demis Hassabis, CEO of Google DeepMind, acknowledged these significant pressures and challenges, stating publicly that researchers and engineers urgently need a large and growing number of computing chips to test new and innovative ideas on a scale large enough to achieve effective results. Although the shortage crisis in high-bandwidth memory supplies from major manufacturers such as Samsung, Micron, and SK Hynix is considered an external factor exacerbating the problem, decisions on internal allocation and distribution of available resources among various departments remain entirely in the hands of company management—a policy that has sparked resentment among many ambitious researchers and fueled their desire for change.
Experts leave to establish new companies and labs
This disruptive dynamic did not stop at mere grumbling; it prompted several prominent scientists with household names in their fields to make decisive decisions to leave their prestigious positions and depart the company entirely to establish their own labs and companies. The company witnessed a wave of impactful departures: Andrew Dai, a veteran expert who spent 14 years leading data and prototype model projects such as Gemini and PaLM 2, left in January 2026 to co-found the new lab Elorian, which raised $55 million in record time. Ioannis Antonoglou, a founding engineer and creator of historic models like AlphaGo and MuZero, also left to co-founder Reflection AI, which raised $2 billion. Similarly, Anna Goldie, former co-lead of AlphaChip development, made a similar move and launched her own company, Recursive Intelligence, in late 2025 with massive funding of $300 million. The stark paradox here is that all these brilliant researchers left a company with capital expenditures of nearly $14 billion in a single quarter, yet they confirmed that they were able to secure computing power much more easily and quickly outside the walls of the tech giant.
Expert assessment of the landscape and resource allocation
Commenting on the nature of the crisis, Oren Etzioni, former CEO of the Allen Institute, described this dynamic as an expected result of administrative policies and the internal market of large corporations, where computing resources are rationed and distributed based on administrative seniority and titles rather than unit economic feasibility and research importance. The parent company did not directly dispute the press reports, pointing instead to the scale of its ongoing investments in infrastructure and emphasizing that resource constraints and supply chain issues represent a common situation faced by the entire industry without exception. This justification is largely true, but what makes this specific company’s situation unique is its successful transformation into the largest supplier of vital infrastructure to its direct competitors, while its pioneering researchers are left standing in long and frustrating queues just to secure enough time to use and operate their own company-owned equipment to ensure innovation and creativity continue.
Frequently asked questions
Question: What is the core reason behind Google DeepMind researchers leaving their company?
Answer: The core reason stems from a severe shortage in access to the computing resources necessary to run their research, due to the company allocating and selling these resources to external clients.
Question: What is the scale of the contractual commitments that restricted the company’s internal resources?
Answer: The company entered into historic commitments to provide infrastructure and computing resources valued at up to $40 billion to Anthropic, in addition to parallel arrangements and contracts to serve Meta’s requirements.
Question: How did the resigned researchers deal with the resource shortage problem after leaving?
Answer: Many researchers established their own technology companies and labs, stating that they were able to secure massive resources and funding with greater flexibility outside the old company environment.
Question: Is this resource crisis limited only to this company or is it a general problem?
Answer: The entire technology sector suffers from computing constraints due to chip shortages, but this company’s situation appears exceptional because it has become a resource supplier to its competitors while its own experts are deprived of them.
Question: How are scarce computing resources typically distributed inside these large companies?
Answer: Expert analyses indicate that major corporations often rely on administrative seniority metrics and bureaucratic decisions for internal resource allocation rather than focusing on actual research feasibility.