How Silicon Valley has become the Compute Cartel

Silicon Valley's AI moment has metastasised into something far more consequential than a mere technology cycle. It is now a capital-intensive arms race dominated by a handful of American hyperscalers, a geopolitical chess match accelerated by Trump's recent overtures to Beijing, and a looming liquidity event for OpenAI cleared by Musk's courtroom defeat. Our tech team's reporting from the Bay Area confirms what the All-In "AI bros" have been arguing for months: this is not another bubble but a compute-driven industrial revolution that will rewire the global economy. The question for our international audience is who controls the wires, and at what cost to the broader market.

The circularity of this infrastructure boom is where the story begins. Frontier labs like OpenAI and Anthropic, now valued at $965 billion after raising $65 billion on infrastructure-heavy growth ambitions, cannot exist without the hyperscalers, yet these cloud landlords are simultaneously their largest customers, creditors, and increasingly their competitors. Alphabet, Amazon, Microsoft, Meta, and Oracle are projected to lift annual capital expenditure from $256 billion in 2024 to roughly $602 billion in 2026, with three-quarters directed toward AI infrastructure. This is not merely spending. It is a hardwiring of economic power. When Amazon, Microsoft, Alphabet, and Meta alone spent $130 billion in Q1 2026, they were not buying capacity for the current cycle; they were cementing a structural advantage that will persist for decades.

The Feudal Reality

The bulls, Chamath Palihapitiya, David Sacks and their fellow podcasting investors, read the mechanism correctly. Token budgets at frontier start-ups now rival payrolls, and the marginal cost of deploying intelligence is sliding toward zero. Their thesis is abundance: a democratisation of capability that lifts productivity across the whole economy.

Jefferies analyst Brent Thill captured the bull camp at its most pugnacious when he told the Financial Times that recent revenue growth justified the outlays and that "the bear thesis is garbage."

Yet abundance and concentration are not opposites here. They are the same trade viewed from opposite ends. The capital flooding into AI is not a rising tide lifting all boats. It is a concentrated suction of liquidity into a few balance sheets, and it is increasingly financed with debt. Big tech has issued tens of billions in bonds in 2026 alone to fund the build-out, and investors have demanded record protection against that debt through credit default swaps. The economy is bifurcating into the AI-capitalised and the capital-starved, and the gap is being underwritten by leverage that simply did not sit on these balance sheets two years ago.

Musk wants to close the loop entirely, and this month he cleared the biggest thing standing in his way. On 18 May a California jury threw out his suit against OpenAI, finding he had waited too long and missed the three-year window to sue, a verdict Judge Yvonne Gonzalez Rogers accepted.

OpenAI's lawyers had dismissed the case as "a hypocritical attempt to sabotage a competitor," and Wedbush's Dan Ives told clients the ruling took "a worst-case scenario off the table" for the company's eventual listing. But Musk has never been one to fight in court alone. He is building: wiring xAI's data centres into SpaceX's launch business, talking openly about compute in orbit and eventual "space fabs" beyond the reach of any single power grid or regulator. It is the hyperscalers' circular logic pushed past the atmosphere.

The Contest for the Wires

The geopolitical stakes radiate outward from there. Washington's AI diplomacy, visible in Trump's Beijing summit and the State Department's planned "digital week" at APEC in Chengdu this July, is a scramble to lock in GPU supply chains and energy contracts that now shape military decision speed as much as commercial advantage. A senior State Department official framed the Chengdu push as a chance to "engage with all 21" APEC economies. Beijing's counter is its own: driving AI deep into its domestic economy at speed while open-sourcing models for the Global South, a reminder that infrastructure has become the primary lever of twenty-first-century influence.

The abundance optimists miss the exclusionary mechanics. No Pacific island state, no Gulf petro-economy and no Atlantic democracy now sits outside this loop. The $500 billion Stargate venture binding OpenAI, SoftBank and Oracle, blessed by the White House, is not merely domestic plumbing. It is the architecture of tomorrow's governance, an arrangement in which sovereignty is increasingly leased from the lords of cloud and silicon.

The Outlook

As the second half of 2026 opens, the contours of this order are no longer ambiguous. The capital is committed, the concrete is curing, and the circularity is self-reinforcing. Expect hyperscaler capex to hold or climb through year-end as enterprise buyers graduate from pilots to full deployment, deepening the divide between the AI-integrated and the left behind. Expect the public listings of Anthropic and OpenAI to crystallise that concentration, funnelling fresh capital back into the terrestrial and orbital loop. And expect Washington and Beijing to keep contesting the single prize that matters more than any product: the right to be the default landlord of global compute.

For markets and governments alike, the comfortable illusion of detachment is dissolving. The second half of 2026 is unlikely to bring correction or consolidation in any ordinary sense. It will bring the consolidation of power itself. The bill is coming due, and the invoice is written in silicon.

Get the stories that matter to you.
Subscribe to Cyber News Centre and update your preferences to follow our Daily 4min Cyber Update, Innovative AI Startups, The AI Diplomat series, or the main Cyber News Centre newsletter — featuring in-depth analysis on major cyber incidents, tech breakthroughs, global policy, and AI developments.