Alibaba Bleeds $10B, Nvidia Gambles on Perplexity: The AI Hardware Crunch No One’s Talking About

(SeaPRwire) –

By: Reginald Vance

Alibaba just raised $10.2 billion in a Hong Kong share sale. This is the largest primary follow-on ever completed by a Hong Kong-listed company. The money goes toward chips, computing infrastructure, and model development. Shares fell more than 8% on the news. Investors are spooked. Not by the AI bet itself. But by the fact that Alibaba’s quarterly profit dropped 75% even before this capital raise. The burn rate is accelerating faster than revenue can keep up. The question on every institutional desk is simple. When does the return on investment show up? There is no answer yet. Here is what the numbers actually tell us. Alibaba is spending heavily on AI infrastructure. They need chips. They need computing power. They need to develop their own models. The $10.2 billion is not discretionary spending. It is survival spending in an arms race. Every cloud provider and tech giant is bidding for the same limited supply of advanced silicon. The capital bottleneck is real. Foundry capacity at leading nodes is constrained. TSMC and Samsung’s own fabs are running at or near maximum utilization. Alibaba’s profit collapse is not a sign of poor strategy. It is a sign of a market where capital expenditure has outpaced revenue generation by a widening margin. The company is buying time. Time to catch up. Time to build infrastructure that may or may not pay off. The broader market is already pricing in doubt. Geopolitical friction is adding to the uncertainty. U.S.-Canada trade talks collapsed. Canada announced retaliatory tariffs starting September 8. The two countries share deeply connected supply chains in automotive and industrial manufacturing. A prolonged dispute could raise costs for companies operating on both sides of the border. For hardware vendors already squeezing margins on capital expenditure, tariff risk is another variable to factor in. The panic is not about AI being a bad bet. It is about the economics of winning an arms race where the ammunition is measured in billions and the supply is finite. Even if Alibaba secures the chips it needs, the question remains whether the revenue model can support the cost structure. The 75% profit drop suggests the answer is no, not yet.

Nvidia is playing a different game even as it faces the same capital pressures. The company is in talks to invest in AI search startup Perplexity as part of a new funding round. The startup is now valued at more than $30 billion, up from roughly $20 billion just one year ago. Perplexity’s annualized revenue jumped from under $250 million at the start of 2026 to over $750 million. That is a 3x growth in a single year. The investment extends Nvidia’s reach into AI software. Beyond its core chip business. But Nvidia shares fell 2.4% Monday. Investors trimmed positions ahead of Wednesday’s earnings report. The broader semiconductor sector took a beating. Marvell and Micron each dropped more than 6%. Sandisk fell more than 10%. Demand for Blackwell chips is being scrutinized. Data-center revenue guidance is what matters. The supply chain story behind these moves is what institutional investors are really pricing in. Nvidia’s Blackwell architecture requires advanced packaging from TSMC. The company is locking in foundry capacity for the next two years. But every AI startup and cloud provider is doing the same thing. The bidding war for HBM memory from Samsung and SK Hynix is driving costs up across the board. Perplexity’s 3x revenue growth is impressive. But the valuation implies expectations that require sustained demand for inference workloads. If Nvidia’s data-center revenue misses guidance, the entire narrative of AI capex paying off comes into question. The sector is one earnings report away from a repricing. The fabrication node yields at TSMC’s leading processes are the real bottleneck. Every company needs the same capacity. The supply is not expanding fast enough to meet the demand curve. This is the physical reality behind the market panic. The Perplexity deal is Nvidia’s attempt to secure demand for its chips from the ground up. By investing in an AI software company, Nvidia is creating a downstream customer that will need its hardware. It is vertical integration by capital allocation. But the strategy only works if Perplexity’s growth trajectory holds. At a $30 billion valuation, the bar is extremely high.

Samsung knows what the pressure feels like. The company announced a shareholder return program worth between 90 trillion and 110 trillion won. Roughly $65 billion to $80 billion. The largest in its history. Yet shares tumbled more than 8%. Investors wanted a more aggressive buyback program. SK Hynix announced a $28.6 billion buyback. The company committed to returning over half of future free cash flow. Samsung’s memory business is critical to AI demand. The market is pricing in a consolidation phase. Only the players with the deepest pockets will survive the capital intensity of next-generation semiconductor production. The real story here is capital allocation under extreme pressure. Samsung is trying to return cash to shareholders while simultaneously investing billions in new fabrication capacity. The gap between what investors want and what management is delivering is widening. SK Hynix is outperforming on buyback aggression. The memory market is cyclical. AI demand is real but uneven. The companies that survive this cycle will fund the next generation of nodes without taking on destructive leverage. Samsung’s $80 billion return program sounds impressive. But it is largely made up of dividends. The buyback component is modest compared to SK Hynix’s commitment. In a sector where foundry economics favor scale, this kind of misalignment signals strategic drift. The hardware vendor consolidation endgame favors players who can fund both R&D and shareholder returns without dilution. Right now, that list is getting shorter every quarter. The capital bottleneck is not going away. The only question is which companies run out of cash first. Samsung’s share price reaction proves that the market no longer believes in dividend promises. It wants buybacks. It wants proof that management can allocate capital under pressure. The gap between the $80 billion return program and the actual buyback commitment tells you everything you need to know about where Samsung stands in this consolidation race. The hardware market does not reward intention. It rewards execution under capital constraints. Right now, the constraint is the bottleneck itself.

Author bio: Reginald Vance, venture partner specializing in semiconductor valuation and advanced materials with over 15 years of experience tracking foundry economics and capital deployment across the Asia-Pacific hardware supply chain.