Semiconductor cash flow triples as Big Tech AI spending hits $1.8T

The streaming and enterprise video ecosystem is witnessing a historic capital transfer from platform operators to hardware suppliers. As hyperscalers burn cash to build the compute foundations for generative AI, high-value chipmakers have secured unprecedented pricing power and liquidity. For video platforms, this suggests that the cost of scaling AI-integrated features—from server-side encoding to personalized recommendation engines—will remain elevated as supply bottlenecks for specialized accelerators persist. Industry watchers must monitor if hyperscalers adjust their subscription pricing or advertising margins to offset this massive capex-driven cash drain. Watch for ASML’s Q2 net bookings as a signal for the 2027 capacity outlook.

The divergence in cash flow dynamics is being intensified by an aggressive upward revision in capital spending plans among the 'Big Four.' Per Business Insider in April 2026, Amazon, Microsoft, Meta, and Google updated their 2026 capex guidance to a combined $725 billion, an increase of roughly $100 billion over previous estimates. Microsoft leads this surge with a projected $190 billion annual spend, while Amazon maintains its eye-watering $200 billion forecast. These figures represent the largest concentrated infrastructure investment cycle in the history of the technology sector, effectively doubling the spending levels seen just two years prior.

This spending is increasingly funneling into specialized silicon and high-bandwidth memory. Per Barron's and Reuters in July 2026, Micron Technology’s recent quarterly revenue surged 345% year-over-year to $41.46 billion, driven by the insatiable demand for HBM3E memory chips used in AI accelerators. While hyperscalers like Google and Meta are attempting to mitigate costs by developing internal custom chips—such as Google’s TPU v6—they remain heavily dependent on external suppliers for merchant silicon and advanced lithography equipment.

Regulatory scrutiny is also mounting as the concentration of capital becomes more pronounced. Financial Times reported in May 2026 that the DOJ and FTC have launched inquiries into the 'acqui-hire' strategies used by Microsoft and Google to bypass traditional M&A reviews. Furthermore, the sheer scale of the buildout is creating secondary bottlenecks in power and cooling. Per Deloitte in June 2025, U.S. data center power demand is projected to grow thirtyfold by 2035, reaching 123 gigawatts. This suggests that even if hardware supply stabilizes, total cost of ownership for AI infrastructure will be pressured by rising utility costs and grid constraints through the end of the decade.