Jensen Huang arrived in Taipei early and dropped the headline: NVIDIA will spend up to $150 billion per year in Taiwan. AMD's Lisa Su followed with a $10B+ commitment. Computex (June 2-5) adds a first-ever robotics zone, drawing Taiwan's full sensor/motor/reducer ecosystem. McKinsey framing: "Taiwan's AI role is moving from a semiconductor story to an infrastructure story." This is the narrative shift from chip fab to full-stack AI systems.
Dell reported revenue of $43.8 billion (nearly doubled YoY), with AI-optimized server orders up 757% to $16.1B. FY27 guidance raised from $138-142B to $165-169B. Stock up ~33% today alone, now up 235% YTD. Dell's COO: "AI opportunity shows no signs of slowing." The company now expects $60B in AI server revenue for FY27, up from $50B prior guidance. This is as loud as a demand-side signal gets for the entire NVIDIA supply chain.
This week's Barron's cover: "The Chip Rally Has Gone Parabolic." SOX up 81% YTD, best first 100 days ever. Micron up 225%. AMD's market cap just passed JPMorgan Chase ($4.3B net income vs JPM's $56.8B). A separate Barron's piece asks outright: "If it walks like a bubble..." BofA survey: 73% of fund managers call "long semis" the most crowded trade. Counter-thesis from Ned Davis Research: this is a supercycle, not a bubble. "Compute power is becoming infrastructure like railroads, utilities, and the internet."
Unitree's STAR Market IPO ($580M raise) goes to listing committee review this Sunday. But the numbers tell a complicated story: Q1 profits down 52% YoY despite revenue growth. 73.6% of humanoid revenue still from research/education customers, not commercial. Patent portfolio limited (262 patents). Meanwhile, Figure AI is producing one robot per hour (350 units) and ran a 24-hour autonomous test. The gap between humanoid hype and commercial reality is becoming the story.
Foxconn Chairman Young Liu declared the company is "accelerating its move into EVs, robotics, and smart cities." Separately, DIGITIMES reports that thermal management is emerging as a critical bottleneck for humanoid robot mass production -- component efficiency, cramped joint architectures, and limited heat dissipation create a physics wall that software can't solve. Also: embodied reasoning is the new frontier, with a DIGITIMES column arguing robots need "deeper thinking" beyond pattern-matched LLM responses.
Barron's cover calls out the quality gap in the chip rally: NVIDIA looks cheap while low-margin names like ON Semi and STMicro are up 122%. Gavin Baker (Atreides): "The multiples cannot all be accurate." Barron's picks: AMD, AVGO, TSM, NVDA.
The Figure AI CEO posted on Threads May 27, garnering engagement. Figure is now producing one robot per hour and just ran a 24-hour autonomous package-sorting test. This is the production milestone conversation you can add to.
Dell is the #1 stock story today. S&P 500 at new records for a 3rd straight day. Dell shares up 33%, joining a select group of stocks up 200%+ in 2026. AI server orders up 757%. Every finance and tech account will be discussing this.
Computex 2026 adds a dedicated robotics zone for the first time. Delta Electronics showcasing "Physical AI Applications." Laster Tech entering robot joint control supply chain. Taiwan supply chain positioning for humanoid components. This bridges AI infra and robotics -- the exact intersection nobody else occupies.
Core thesis: Computex just added a robotics zone for the first time. The same Taiwanese companies that build AI server racks -- Foxconn, Delta Electronics, Pegatron -- are now building humanoid robot components. Jensen Huang's $150B/year Taiwan bet isn't just about GPUs. It's about the physical AI stack: from data center to robot body. The compute layer and the embodiment layer are converging in the same supply chain, and Taiwan is the bottleneck for both.
Why this works: Nobody in AI Twitter connects the semiconductor supply chain to the robotics supply chain. You're the person tracking both. This is a 3-4 post thread, not a 15-tweet essay. Keep it tight.
Core thesis: Unitree files for a $580M IPO. Revenue is growing. Profits are shrinking. 73% of humanoid revenue comes from researchers, not factories. Meanwhile, Figure AI just ran a 24-hour autonomous test. The question isn't "are humanoid robots the future?" -- it's "when does the revenue shift from labs to loading docks?" Unitree's prospectus is the most honest document in robotics right now because it shows both the promise and the gap.
Why this works: Contrarian but informed. You're not dunking on robotics -- you're showing you read the filings and can separate hype from traction.
Core thesis: Dell's AI server orders up 757%. Revenue guidance raised by $25B+. The "is AI real?" debate is over at the infrastructure layer. The new question is who captures value at each layer of the stack: chips (NVIDIA), fab (TSMC), servers (Dell), inference (the cloud), and the output layer where agents actually deliver work. As someone building at the output layer, watching Dell's numbers validates the entire chain underneath.
Why this works: Timely reaction to the biggest earnings story of the day, but with a builder's perspective, not just an investor's. Post within hours for maximum relevance.
Threads is showing strong engagement on robotics and AI investing content this week. A robotics-adjacent post pulled 7,550 likes and 325 comments (May 28). Brett Adcock's Figure AI updates are getting meaningful engagement. Semiconductor and chip investing content is also active with multiple accounts posting NVIDIA/TSM analysis.
The Barron's bubble piece and the Ned Davis Research supercycle counter-thesis create a natural engagement hook. On Threads, the format is simple: pick a side, state why, in 3 sentences. "73% of fund managers say long semis is the most crowded trade. But compute is becoming infrastructure. The real question isn't whether it's crowded -- it's whether the demand curve bends."