2 · Today's Digital Landscape
The boom is real. So are the bottlenecks.
Thursday's thread is asymmetry: capital, chips and compute can scale frighteningly quickly; embodied intelligence still has to survive friction, messy environments and customers who expect the machine to work every time.
NvidiaAI infrastructureDemand
Nvidia just told the market the AI build-out still has another gear
Nvidia's latest forecast helped push technology shares higher after the company projected roughly 70% revenue growth for the fiscal year ending in early 2028. That is not a model benchmark or a startup valuation; it is a supplier telling investors that customers still intend to buy an extraordinary amount of AI compute. The caveat is sitting inside the same story: memory supply remains constrained. The AI boom keeps proving both halves of the infrastructure thesis at once — demand is enormous, and every layer underneath the GPU eventually gets a turn at becoming the bottleneck.
Source: Reuters ↗
Humanoid robotsChinaReality check
China can manufacture the humanoids. Making them useful is taking longer.
Reuters' investigation into China's humanoid industry lands like a bucket of cold water on several weeks of robot headlines. China produced the overwhelming majority of the world's roughly 20,000 humanoid shipments last year and has poured subsidies into the sector, yet many systems remain slow, brittle and dependent on carefully structured routines. Traditional industrial robots still beat them at the repetitive work factories actually pay for. The important correction is not that humanoids are fake. It is that manufacturing scale and embodied intelligence are different problems. You can subsidize factories into existence faster than you can subsidize dexterity, generalization and reliability.
Source: Reuters ↗
AnthropicComputeNscale
Anthropic reportedly bought six years of future compute for about $45 billion
Anthropic has reportedly agreed to rent roughly $45 billion worth of AI compute from British infrastructure company Nscale over six years, with capacity expected to begin coming online in late 2027 using Nvidia's Vera Rubin systems. TechCrunch says the deal follows a string of other enormous compute commitments from Anthropic involving Volta, AMD, Amazon, Google, Broadcom and SpaceX. The number is useful because it turns an abstract arms race into a procurement schedule. Frontier labs are no longer merely buying GPUs; they are reserving pieces of the future electricity-and-silicon economy years before the workloads arrive.
Source: TechCrunch ↗
