Nebius' software efficiency and a possible end to heavy AI financing support a re-rating. Per @DrTomsLens, its engineers got about 43% more video output from the same GPUs, cutting compute cost per second by 30%; newer hardware alone is not enough without capable software teams. Oracle expects to fund growth from free cash flow in FY2029, suggesting capital-heavy AI providers may eventually outgrow debt and dilution. Google Cloud reported 103% revenue growth amid strong AI infrastructure demand. … Show more

Cheaper inference multiplies compute demand for Nebius, Cipher Mining and IREN rather than reducing it, because more people can afford to run bigger workloads. That is Jevons paradox playing out in real time, not a bear case from Chinese open-weight models. x.com 

AI hyperscalers' heavy bond issuance can crowd out the corporate AI spending they depend on. After hyperscalers flooded the bond market all summer, a company rolling its Covid-era debt 5% higher has less left for an AI budget. And if every valuation and megatech management guide rests on commitments from Anthropic and OpenAI, they rest on unaudited financials never reported to the SEC. x.com  x.com 

Neoclouds such as Nebius, CoreWeave, IREN and TeraWulf remain long-term winners because cheaper, better AI models drive more compute use, not less. Per early testing by @Signal_65, GPT-6 Astra completed 279 of 280 enterprise tasks with zero fabrication, cutting weighted errors about 51% versus Claude Fable 5.1 at lower cost. Enterprises are deploying far more agents, pushing the upper end of 2028 capex estimates for the five biggest hyperscalers toward $1 trillion. x.com 

Financing announcements point to more accelerator demand, supporting semiconductors. About $200bn in announced Meta, Anthropic/Broadcom and SpaceX/Nvidia financing likely drives $400bn in XPU purchases. Data centers burn cash until lit, then cash flow 20-40% annually; AI spend rising from 0.1% to 0.5-1% of current payroll could balance the buildout if agents replace roles. Sustainable token demand is the bet, but smaller local models are the bigger risk, though not an immediate concern. x.com  … Show more

Meta's hundreds of billions in AI investment have produced Muse, a free, largely undifferentiated personal assistant app that will likely never generate economics. Muse topped app-store downloads and helped fuel a two-day gamma squeeze mostly because it is new and free. Amazon has already cut it off from shopping its site, and once the novelty fades it will probably see little use given its limited utility; people actually like doing their own shopping. x.com 

Inference demand, not model release cadence, drives AI infrastructure, so dips in inference suppliers are unwarranted. Anthropic's CEO, with OpenAI's Sam Altman agreeing, committed only to third-party evaluations before releases, saying pacing does not mean halting training. Even if releases slowed, inference demand keeps climbing and older models need more infrastructure, benefiting Nvidia's GPUs, Micron's memory, Astera Labs' GPU connectivity, Marvell and Broadcom. x.com  x.com  x.com 

Arm's data-center case rests on fleet economics for agentic AI, not single-chip benchmarks. Per Arm's cloud EVP Mohamed Awad, GPUs generate tokens and CPUs decide what to do with them, so more tokens need more CPUs. Thousands of sequential agent threads running at once reward fast cores, memory bandwidth per watt and efficient scaling across a data center, where x86-era CPUs fall short. Meta is Arm's first customer, with OpenAI and Cerebras also on board. x.com  x.com  x.com 

Meta's rally on its Muse AI may have bought the wrong layer, because a billion users is not a hardware spec. Modeling the workload, the token bill and who gets paid to let agents through points elsewhere: Wall Street bought CPUs, yet agents still need a human. x.com 

AI agents scaling to hundreds of millions of users will need a large layer of infrastructure, and the buildout is still early. Agents need CPUs to run browsers, code, tools and background tasks, adding a major new source of server CPU demand for Intel and AMD. They also require more memory, compute and data center capacity, benefiting neoclouds like IREN and Nebius. Underneath it all, more data centers mean more electricity. x.com