🚀 Cerebras Rockets to $64B, Walsh Confirmed as Fed Chair, and Musk v. OpenAI Wraps Up

💥 Cerebras IPO Crushes Expectations

The Cerebras IPO delivered one of the most spectacular debuts in recent memory, with the AI chip designer's market cap reaching $64 billion on its first day of public trading. Trading at $300 per share (after briefly touching $350), the company's valuation journey tells a remarkable story of consistent execution and market validation.

The IPO pricing itself underwent multiple revisions upward — starting from an initial range of $115-$125, then moving to $150-$160 on Monday, before ultimately settling higher. Prediction markets hadn't even included a category above $50 billion, with most participants betting well below that threshold. One-third of investors who requested shares in the order book received zero allocation, while the top 25 institutional investors captured 60% of the offering.

"This IPO illustrates the power of an individual partner over the brand name of the firm. Pierre Lamond was a partner at both Sequoia and Kleiner, but instead of those firms backing Cerebras, it was Eclipse, the firm he joined at the age of 84 that backed this little-known chip company multiple times in the early days. What a way to wrap up a career. He was born in 1930, the same year as Warren Buffett." — Honam

🔬 The Technology Behind the Hype

For those unfamiliar with Cerebras' approach: instead of cutting a silicon wafer into dozens of smaller chips, the company uses the entire wafer as a single massive chip. This audacious design choice faced early skepticism around defect rates and yields, but Cerebras solved these challenges by building in redundant cores that aren't all activated simultaneously.

The chips are already serving production workloads. OpenAI uses Cerebras chips to power Codex 5.3 Spark, and the company has secured a massive 750-megawatt deal with OpenAI for future capacity. The user experience is distinctly different from traditional LLM interactions — rather than watching tokens stream in one by one, responses appear almost instantaneously, like loading a complete Wikipedia page.

⚡ The Speed Premium: Why Customers Pay 6x for 2x Performance

Perhaps the most revealing insight from Semi Analysis' coverage: customers are demonstrably willing to pay disproportionately more for faster inference. Anthropic's Opus 4.6 fast mode charges six times the price for roughly two times the speed compared to standard mode. Despite this seemingly poor value proposition, Semi Analysis reported spending 80% of their AI budget on the fast mode.

This revealed preference contradicts earlier assumptions that users would prioritize raw intelligence over interactivity. The analogy to e-commerce latency is instructive: Amazon famously discovered that every 100 milliseconds of page load time cost them approximately 1% in sales. Similarly, LLM users who fire off a query and then wait often get distracted, scroll social media, and forget to return to their original question.

In business contexts, the value of speed compounds further. An employee who produces the same quality work but operates twice as fast doesn't command twice the salary — they often command five or six times the compensation due to their multiplicative impact on organizational output.

⚠️ Technical Headwinds on the Horizon

Despite the successful IPO, Semi Analysis identified several meaningful challenges ahead:

• Memory limitations: Cerebras chips currently struggle with larger models and limited on-chip memory compared to networked GPU clusters like Nvidia's NVL72 racks
• Context window constraints: The industry is trending toward larger context windows, but Cerebras' current architecture supports only 128K tokens — likely insufficient for emerging agentic workloads
• SRAM scaling problems: The WSE-2 chip had 40GB of memory; the WSE-3 increased this to only 44GB — a mere 10% improvement over one full process node iteration
• Architectural trade-offs: Adding more SRAM requires sacrificing compute area since everything must fit on a single wafer

The path forward likely involves hybrid orchestration architectures. Large, powerful "boss" models running on traditional GPU clusters could delegate specific tasks to smaller, faster Cerebras-powered workers — similar to how current LLMs already offload database queries and web searches to CPUs.

📈 The Valuation Journey

Cerebras' funding history reflects steady conviction from early backers:

• Series A (2016): $100M valuation — Foundation, Benchmark, and Eclipse
• Series B (2016): Undisclosed valuation — Coatue Capital led
• Series C (2017): Undisclosed valuation
• 2018 round: $1.6B valuation
• 2019 round: $2.4B valuation
• 2021 round: $4B valuation
• 2025 round: $8B valuation — A16Z and Fidelity
• Late 2025: $23B valuation — Tiger Global
• May 2026 IPO: $48.8B initial pricing, closing day one at $64B market cap

🏦 Kevin Walsh Confirmed as Fed Chair

In a largely party-line vote, the Senate confirmed Kevin Walsh as the 17th Federal Reserve Chair with a tally of 54-45. The narrow margin stands in stark contrast to Jerome Powell's confirmations, which captured at least 80 votes in each of his two terms.

Walsh inherits a challenging macroeconomic environment potentially characterized by stagflation — simultaneous economic stagnation and inflation. This scenario presents a particular policy dilemma: raising rates to combat inflation risks deepening economic weakness, while cutting rates to stimulate growth could fuel further price increases.

The divided Senate vote signals potential friction ahead, particularly given public pressure from the White House for rate cuts amid persistent inflation concerns. Walsh's nomination attracted attention partly due to his memorable interview with Palantir CEO Alex Karp, during which Karp appeared to have a nicotine pouch in his mouth while spinning a notebook on his finger throughout the conversation.

⚖️ Musk v. OpenAI: Closing Arguments and High Stakes

The blockbuster trial between Elon Musk and OpenAI entered its final day, with both sides delivering closing arguments. The case, which many expected to run four weeks, concluded in just three.

The Stakes: Musk is seeking more than $150 billion in damages and requesting the court to remove Sam Altman from OpenAI's board while blocking the company's conversion to a for-profit structure.

Prediction markets initially gave Musk a 58% chance of winning in late April, but that probability has since collapsed to just 30% as of the closing arguments.

Key Arguments:

Musk's legal team focused on:

• Portraying Sam Altman as fundamentally untrustworthy
• Arguing that OpenAI executives are enriching themselves while operating as a supposed charity
• Emphasizing Musk's concern for humanity versus alleged greed at OpenAI
• Attacking witness credibility through repeated character challenges

OpenAI's defense centered on:

• Asserting that "even people who work for him, even the mother of his children can't back his story"
• Arguing there was never a specific, legally binding agreement between Musk and OpenAI regarding how his donations should be spent
• Characterizing Musk's claims as legally unsupportable regardless of the surrounding narrative

"Not one person in this case other than Elon Musk has testified to any commitments or promises that Sam Altman or Greg Brockman or OpenAI made to Mr. Musk." — Sarah Eddie, OpenAI legal team

The trial featured its share of courtroom drama, including a dispute over a large monitor brought in by Musk's legal team. After OpenAI's lawyers requested to use it, Musk's team initially refused before the judge intervened. Approximately 50 lawyers then crowded the courtroom floor attempting to solve the technical challenge of connecting laptops to the display.

Notably, Musk was not present for closing arguments — he was in China accompanying President Trump. The jury is now deliberating, with a verdict expected in the coming days.

🤖 Humanoid Robots and the Teleoperation Question

Figure AI's Brett Adcock livestreamed what he claimed was a fully autonomous eight-hour shift of the Helix 2 humanoid robot working at human performance levels. The stream attracted 3.4 million views and initially impressed observers with the robot's speed and dexterity.

However, skepticism emerged when the robot began exhibiting odd behaviors — reaching across its body in ways that seemed inefficient, and at one point reaching up to touch its own head with no apparent logical purpose. Critics questioned whether the system was actually teleoperated by a human wearing a VR headset.

Adcock maintains the system is "fully autonomous," though Figure AI's own promotional materials humorously depicted their "in-house neural network running entirely on board" as a human in a VR headset. The debate highlights ongoing questions about what constitutes true autonomy versus sophisticated human-in-the-loop systems in robotics demonstrations.

🌐 Other Notable Developments

Jensen Huang in China: The Nvidia CEO was photographed in China carrying two large boxes of GeForce RTX 5090 graphics cards, continuing his aggressive go-to-market strategy amid evolving US-China technology export restrictions.

Tim Draper's Cold Plunge Marathon: The billionaire venture capitalist claimed a personal record by taking 52 pitch meetings in 52 minutes while sitting in a 40°F ice bath. The setup, photographed in what appeared to be a makeshift facility with trash bags and exposed piping, drew attention for its surprisingly spartan aesthetic.

For more macro and markets analysis, visit TBPN.com