The Great Divergence Is Here — And Nobody Wants to Say It Out Loud
On the same day Anthropic reportedly plans $750 billion in computing expenditure, Barclays announced it would cut 5,000 jobs in an "AI-fuelled push to lower costs." Qualcomm is shedding headcount after annual reviews. Tesla is cutting spending even as Musk bets the company's future on AI and robots. Read together, these aren't separate stories. They are one story — and the most important economic narrative of 2026 is being told in a way that obscures its true meaning.
The AI revolution is not a clean displacement story. It is a capital allocation crisis masquerading as a productivity miracle. The companies spending trillions (Anthropic, OpenAI, the hyperscalers) are doing so because they believe the returns justify the outlay. The companies cutting 5,000 jobs (Barclays) are doing so because they've decided the returns don't justify the human infrastructure that used to deliver them. Both positions can be simultaneously rational — and both are creating a world where wealth concentrates in fewer hands while labor income stagnates for millions.
This isn't a theory. It's a balance sheet.
The $750 Billion Elephant in the Room
Let's be honest about what Anthropic's reported $750 billion computing commitment means. It represents a capital expenditure decision that exceeds the entire annual GDP of countries like Portugal or the Czech Republic. For context, the entire Manhattan Project cost approximately $2.5 billion in 2026 dollars. We are now funding AI infrastructure at 300 times the scale of the most consequential scientific project in human history — and doing so largely in private, with no public mandate, no congressional oversight, and no mechanism for socializing the losses if the bet fails.
What makes this particularly dangerous is the concentration risk. Anthropic, OpenAI, and a handful of others are making near-binary bets on AGI timelines. If the technology works at the pace they expect, they will own the most valuable assets in economic history. If it doesn't — if we hit a capability wall, a cost wall, or a regulatory wall — we will have a private-sector capital misallocation on a scale that makes the dot-com bust look like a rounding error.
The DeepSeek angle adds a geopolitical dimension that most Western coverage conveniently omits. DeepSeek's new open-source tool is designed to reduce deployment costs on both Nvidia's Blackwell GPUs and Huawei's Ascend 950 chips. That second name matters enormously. Every dollar of AI compute that runs on Huawei hardware is a dollar that bypasses US export controls. DeepSeek isn't just competing on model quality — it's building the software layer that makes Chinese semiconductor independence viable. This is not a startup story. This is a national security story wearing a developer-tools costume.
Asia's Monetary Fracture: The Signal Behind the Noise
The Philippines' 50 basis point surprise hike — its fourth consecutive increase, lifting 2027 inflation forecasts to 6% — tells you everything about how fragmented Asian monetary policy has become. The BSP isn't hiking because the Fed is hiking. It's hiking because El Niño is destroying rice yields and oil prices remain sticky. The drivers are idiosyncratic, not systemic.
Meanwhile, the Bank of Korea holds at 3% with a hawkish tilt, and Indonesia keeps its main rate unchanged while hiking the deposit facility rate and cutting reserve requirements — a configuration that screams "we want to support growth without weakening the currency." Singapore's dollar, meanwhile, is being used as a carry trade funding vehicle against the yen, benefiting from low volatility in a region where volatility is otherwise rising.
This is not coordination. This is each central bank fighting its own war with the weapons it has. The old model — where the Fed moves and everyone else adjusts in lockstep — is dead. What replaces it is a patchwork of responses that will create currency volatility, capital flow disruptions, and a persistent bid for safe-haven assets in the region. Singapore benefits. The Philippines doesn't. Indonesia plays both sides. This is the new normal, and portfolio managers who haven't adjusted their models for it are going to get hurt.
Southeast Asia's Ambition-Constraint Paradox
Taiwan blocking Grab's $600 million Foodpanda acquisition is a small news item with enormous symbolic weight. This was Grab's first attempt to expand outside Southeast Asia. It was blocked. The message to every Southeast Asian entrepreneur reading the HSBC Global Entrepreneurial Wealth Report — which found that the global share of entrepreneurs wanting to expand internationally fell six percentage points — is clear: the world is closing its doors, and even friendly democracies will protect their markets.
Yet simultaneously, we see OCBC adding the Malaysian ringgit to Singapore savings accounts, DBS PayLah expanding QR payments to Taiwan, and Airwallex deepening its McLaren partnership. The financial rails are expanding even as the corporate rails are being blocked. This is the defining contradiction of Southeast Asian growth in 2026: you can move money across borders more easily than you can move businesses. The regulatory perimeter is firming up exactly where it matters most — ownership, control, market access — while fintech quietly builds the plumbing for a borderless financial future that may never be politically permitted to arrive.
Jio's $3.2 billion IPO — an offer for sale with no new capital raised — tells yet another story. This is an exit, not an entry. The existing investors (including, presumably, the Facebook/Meta legacy) are cashing out. In a market where international expansion is being blocked and domestic regulation is tightening (see: India's largest bank just leaking customer records), selling to the public is the rational move. Not because the company is at a peak, but because the optionality of staying private has collapsed.
The AI Jobless Recovery Isn't Coming — It's Already Here
Here's the irony that should keep policymakers awake at night: Barclays is cutting 5,000 jobs because of AI, while simultaneously AI companies are raising hundreds of millions (Otter.ai with $320 million, Blackwall with $100 million, ASI with $325 million) to build the tools that will cut the next 50,000. The AI infrastructure layer is hiring. The AI application layer is displacing. And the companies in between — banks, insurers, logistics firms, automakers — are in the brutal middle ground of having to fund the transition while delivering returns to shareholders who have no patience for it.
Tesla's spending cuts before a potential SpaceX deal, while simultaneously defending a flying Roadster announcement that everyone knows is theater, captures the absurdity perfectly. The market rewards AI narrative as much as AI execution. That's not investing — that's performance art with a P/E ratio.
The Bottom Line
The AI revolution's first act was about capability. Its second act — the one we're living through right now — is about who bears the cost of the transition. The answer, so far: workers at banks and telecoms who never asked to be disrupted, shareholders who are funding $750 billion bets they don't fully understand, and governments in the Philippines and Indonesia who are fighting inflation with one hand while watching their companies get acquired by AI-funded competitors with the other. The revolution will not be televised. It will be downsized, restructured, and filed under "efficiency gains" in a quarterly earnings call.