![\"The](\"https:\/\/s.yimg.com\/lo\/mysterio\/api\/D73902CCC3F269C7B6CDB7BE7543E2BD1D834DFF713843E583DCF1713251E611\/subgraphmysterio\/resizefit_w960;quality_80;format_webp\/https:%2F%2Fmedia.zenfs.com%2Fen%2Finsidermonkey.com%2F6e05024de04e2f40c1093571a18bf29f.jpg\")
<\/a> The Trillion-Dollar Token Crash: Why Too Much AI Hardware Could Be Nvidia\u2019s Trap and Microsoft\u2019s Goldmine Too Much Capacity, Too Soon <\/p>\nSkeptics think the industry is building hardware capacity faster than customers can use it profitably. Under the bear case, the revenue gap eventually reaches hardware suppliers. If enterprise demand fails to fill new capacity, hyperscalers will slow purchases, demand lower prices and move routine workloads to internal chips. Short hardware lifecycles would make even a temporary glut expensive. <\/p>\n
Nvidia Corporation (NASDAQ:NVDA<\/a>) carries the greatest exposure because its 74.9% quarterly gross margin depends on customers competing for scarce, high-end GPUs. Google’s TPUs, Amazon’s Trainium and Microsoft’s Maia can absorb predictable inference workloads, reducing Nvidia purchases and increasing hyperscalers’ bargaining power. <\/p>\n The market still expects Nvidia Corporation (NASDAQ:NVDA) to retain most of its advantage. The shares traded at 25.64 times forward earnings as of August 17, while 285 elite hedge funds in Insider Monkey’s second-quarter database held long positions. Slower orders combined with weaker pricing would hit both earnings expectations and a crowded trade. <\/p>\n Cheaper AI Creates More Uses <\/p>\n Bulls focus on the other side of the equation: using AI is getting much cheaper. GPT-4-level inference has fallen from roughly $20 per million tokens to about $0.40, a 50-fold decline. Deloitte estimates that inference will account for roughly two-thirds of AI compute in 2026. <\/p>\n Lower costs do not necessarily mean lower demand, however. An AI agent may call a model repeatedly to search, plan and check its work. As those calls get cheaper, developers can use far more of them. Falling unit costs could expand the market, as they did in cloud computing. <\/p>\n Microsoft (NASDAQ:MSFT) shows how that can become durable revenue. Its AI operations exceeded a $37 billion annual run rate in the March quarter, up 123% from a year earlier, while Microsoft 365 Copilot passed 30 million paid seats by June. The company can spread cheaper inference across Office, Azure, GitHub and Dynamics. Its advantage is not simply the model, but its place inside software companies already use. Hedge-fund support remained broad but weakened as rising AI capex, depreciation and weaker free-cash-flow conversion threatened to outrun Copilot and Azure revenue: 275 funds in Insider Monkey’s database held the stock at end of Q2, down from 287 a quarter earlier. Short interest fell 12% in the second half of July to 1.10% of the float, showing that most skeptics trimmed long exposure instead of building bearish positions. <\/p>\n The Most Boring Outcome is the Most Likely <\/p>\n An outright dot-com style crash is unlikely, as is a flawless revenue climb to $1 trillion. The most realistic outcome is probably a staged infrastructure digestion phase. <\/p>\n The adjustment could gradually shift value from hardware scarcity toward software monetization. As inference becomes cheaper and custom silicon absorbs routine workloads, Nvidia could face slower growth and more cyclical margins despite its platform leadership. Microsoft (NASDAQ:MSFT<\/a>) would likely capture more of the upside by embedding cheap inference into products customers already pay for, converting the infrastructure buildout into recurring revenue after absorbing the near-term depreciation and cash-flow pressure. <\/p>\n While we acknowledge the potential of MSFT and NVDA as investments, we believe certain AI stocks offer greater upside potential and carry less downside risk. If you’re looking for an extremely undervalued AI stock that also stands to benefit significantly from Trump-era tariffs and the onshoring trend, see our free report on the <\/a>best short-term AI stock<\/a>. <\/p>\n READ NEXT: NVIDIA (NVDA): What Foxconn and Super Micro Are Telling Us about the AI Boom<\/a>\u00a0 and Pony AI Is Scaling Robotaxis Fast\u2014Can the Stock Reach BofA’s $17 Target?<\/a> <\/p>\n