AI Broadened. Rates Got in the Way.
GOOD MORNING, Wall Street lost momentum Thursday as oil jumped almost 4%, Treasury yields rose and investors waited for Friday's jobs report. The macro setup was simple: higher energy costs keep inflation risk alive, while higher yields raise the hurdle for expensive growth stocks. But after the bell, Cloudflare showed why the AI trade remains difficult to fade. Revenue grew 36%, guidance moved higher and AI agents are driving more traffic through its network. At the same time, SpaceX and Tesla committed $16.8 billion to build their own advanced chip complex in Texas. AI demand is broadening across software and physical infrastructure — just as the cost of financing it is getting harder. MARKETS | TLDR
AI MONETIZATIONCloudflare Is Turning AI Agents Into RevenueCloudflare reported second-quarter revenue of $696.1 million, up 36% from a year earlier, and raised its full-year revenue forecast to $2.86 billion to $2.87 billion. The company also lifted adjusted EPS guidance and said current remaining performance obligations grew 35%. Shares jumped about 18% after hours as investors focused on the speed at which AI-driven demand is moving through the business. That matters because Cloudflare sits well outside the GPU layer. AI agents and machine-to-machine workloads create more API calls, more inference traffic and more security requirements, all of which increase demand for Cloudflare's network, Workers developer platform and security products. The company added roughly 2 million developers during the quarter — more than it added in all of last year — making the results a concrete sign that AI monetization is spreading into recurring software and infrastructure revenue. The caveat is valuation and execution. Cloudflare remains expensive relative to most software peers, and the company is still absorbing the costs of restructuring while scaling usage-based products. But the broader signal is important: the AI trade is no longer only about who sells the accelerators. It is increasingly about who gets paid every time those accelerators generate traffic. PHYSICAL AISpaceX and Tesla Are Building Their Own Chip Bottleneck AwaySpaceX and Tesla said they will initially invest $16.8 billion to build Terafab, an advanced semiconductor complex in Grimes County, Texas. The companies say their combined future compute needs could exceed 1 terawatt, and SpaceX described existing and planned global chip supply as insufficient for the scale of AI, robotics and space-compute workloads they expect to run. The project matters because it takes the AI-capex cycle one step further upstream. Companies started by buying GPUs and leasing data centers. Now some large users are trying to control logic chips, memory and advanced packaging themselves. Terafab is a bet that future compute demand becomes large enough to justify owning semiconductor manufacturing capacity rather than relying exclusively on TSMC, Samsung or Intel. The economics are much harder than the strategy sounds. Semiconductor fabs require enormous upfront capital, specialized equipment, power, water and years of process development before output arrives. Future phases could push the project's investment far above the initial $16.8 billion. The takeaway is not that vertical integration will automatically work; it is that AI demand has become large enough for customers to consider spending tens of billions of dollars just to secure supply. HEADLINES
UPCOMING
DEEP INSIGTHSCloudflare Q2 2026 ResultsThe primary release is useful because it shows AI monetization at a different layer of the stack. Revenue accelerated to 36% growth, current RPO rose 35% and developer adoption expanded sharply, providing evidence that AI agents are creating recurring demand for network, security and edge-compute services. SpaceX and Tesla's Terafab PlanRead this for the physical scale now embedded in AI forecasts. The companies are starting with $16.8 billion because they believe future compute requirements could exceed 1 terawatt. Whether or not that forecast proves accurate, it shows how AI investment is moving from chips and cloud contracts toward fabs, utilities and industrial infrastructure. |