The artificial intelligence explosion has pushed traditional rivals in Silicon Valley into unprecedented, often uncomfortable, collaborations, forming a complex web of trillion-dollar deals and strategic dependencies that redefine competition and raise significant long-term risks.
The saying “keep your friends close, and your trillion-dollar AI enemies closer” has never been more relevant than in the current artificial intelligence arms race. Silicon Valley is witnessing an intensity of competition not seen since the smartphone boom, as tech titans pour billions into securing the most advanced models, unparalleled compute power, and top-tier researchers.
This fierce competition has paradoxically driven rivals closer than ever, fostering relationships that would have been unthinkable just a few years ago. The stakes are astronomically high, pushing companies into pragmatic, yet strategically perilous, alliances.
The Uncomfortable Embrace: Rivals Becoming Reliances
Recent weeks have revealed the extent of these complex interdependencies:
- OpenAI, majority-backed by Microsoft, inked a $300 billion deal for Oracle’s compute power.
- Meta signed a $10 billion deal for Google Cloud services, according to a person familiar with the matter.
- Microsoft now offers customers access to Anthropic’s AI models, which operate on Amazon and Google’s cloud infrastructure.
- Nvidia made a $100 billion investment in OpenAI, with plans for OpenAI to build at least 10 gigawatts of AI data centers utilizing Nvidia chips. This deal raised concerns reminiscent of Cisco’s vendor financing in the 90s, which historically did not end well.
These arrangements are more than just business deals; they are forming deep and consequential dependencies. As Gil Luria, managing director at investment firm D.A. Davidson, noted, “The stakes are so high that you’re seeing behavior that in the past wouldn’t happen.” It’s a delicate “chess match” of advancing one’s own progress without missing out if a competitor takes the lead.
A Historical Pattern of Co-opetition
Silicon Valley’s history is replete with examples of rivals becoming partners. Consider Google’s long-standing deal with Apple to be the default search engine on the iPhone, a lucrative arrangement that recently drew the attention of the Justice Department. Similarly, cloud giants like Amazon host services from rivals (e.g., Netflix), and Apple relies on Samsung components for its competing phones.
The AI boom has only accelerated this trend, creating new partnerships out of necessity and deepening existing ones. Companies like OpenAI and Meta publicly discuss their reliance on Google Cloud, and Apple has even reportedly trained its AI on Google’s Tensor Processing Units. Even Google itself offers Nvidia’s competing GPUs to its cloud customers, highlighting the intricate web of collaboration.
Rishi Jaluria, an analyst at RBC, points out the pragmatic nature of these deals: “People recognize it’s hard to build large language models, and not only hard, it’s really expensive.” These collaborations allow companies to benefit from the AI boom without shouldering the entire financial burden on their balance sheets. Many also fear missing out, haunted by the specter of past tech bubbles where companies like Sears and BlackBerry failed to adapt.
Strategic Risks and the Data Dilemma
While these alliances offer immediate benefits, they carry significant longer-term strategic risks. OpenAI, for instance, is a major customer for cloud providers like Google and Microsoft, but in doing so, it’s also gaining expertise in building its own data centers, potentially threatening the cloud giants’ business down the line. It’s a “David and Goliath” scenario where the giant is inadvertently helping David forge his slingshot.
The demand for data to train AI models has reached a critical point. Companies like OpenAI, Google, and Meta have reportedly cut corners, ignored corporate policies, and debated bending the law to acquire the vast amounts of digital data needed. OpenAI, facing a data shortage, developed a speech recognition tool called Whisper to transcribe over 1 million hours of YouTube videos, despite YouTube’s prohibition on such use. Similarly, Google transcribed YouTube videos for its own AI models and broadened its terms of service to access publicly available Google Docs and other online material, as reported by The New York Times.
Meta also faced a data crunch, with discussions among managers, lawyers, and engineers about potentially buying publishing houses like Simon & Schuster or gathering copyrighted data without licenses. As Sy Damle, a lawyer representing Andreessen Horowitz, stated, “The only practical way for these tools to exist is if they can be trained on massive amounts of data without having to license that data.”
The ‘Round-Tripping’ Phenomenon and Its Critics
The tangled web extends to financial arrangements, particularly the “round-tripping” phenomenon common in the cloud wars. This involves Company A investing in Company B, which then pays Company A for services, allowing Company A to recoup its investment and demonstrate demand for its services. Examples include:
- Amazon investing $4 billion in Anthropic, which subsequently chose Amazon Web Services as its primary cloud provider.
- Google also being an investor in Anthropic, tying its fortunes to a rival’s performance.
- Nvidia selling chips to smaller cloud providers it invests in, and then reportedly renting those chips back, boosting the startup’s revenue and Nvidia’s stake.
These arrangements are viewed with skepticism by investors because they can inflate numbers and blur the lines between organic revenue and recirculated funds. Gil Luria highlights Nvidia-backed cloud company Coreweave as a “glaring example of bad behavior,” where Nvidia both seeded the company to create competition and then signed up as a customer, ultimately buying back unsold cloud capacity in a deal that seemed “value destructive.”
Is a Bubble Brewing? The Demand vs. Investment Debate
Despite the complexities and potential pitfalls, analysts like Gil Luria believe the demand for AI is “very real” and not going away, unlike past fads like the metaverse. This suggests that if a bubble were to burst, it might only affect the “unhealthy” players. From a Keynesian perspective, if genuine customer demand exists, then large players like Nvidia and Oracle “priming the pump” to stimulate the wider AI economy might not be inherently negative.
Nvidia CEO Jensen Huang articulated this vision, stating to CNBC that companies are building “a brand-new industry called AI infrastructure.” The key, however, is whether ultimate customer demand can sustain this intricately woven infrastructure, valued at over $1 trillion and growing.
Much hinges on figures like Sam Altman, who has positioned OpenAI at the nexus of these deals. As Bernstein analyst Stacy Rasgon ominously put it, “Sama has the power to crash the global economy for a decade or take us all to the promised land.” The future of this trillion-dollar tango remains uncertain, a high-stakes dance with immense potential and equally immense risks.
Hugh Langley is a senior correspondent at Business Insider where he writes about Google, tech, and wealth.
Read the original article on Business Insider
