AI Chips

Alphabet’s custom-designed TPUs are bolstering Google’s AI capabilities and fueling its cloud dominance. These proprietary chips power Gemini and offer significant cost advantages over competitors like Nvidia’s GPUs, making them attractive to AI startups and enterprises. This strategic hardware development positions Alphabet as a formidable player in AI infrastructure, driving strong growth in its cloud division and beyond.

SK Hynix shares surged over 11% after announcing plans for a U.S. stock market listing of American Depositary Receipts (ADRs). This move, expected to raise nearly $30 billion, aims to broaden its investor base and enhance its valuation, particularly in light of booming AI chip demand. The company is significantly investing in U.S. and South Korean facilities to meet this demand, with analysts viewing the listing as a key step to compete with rivals like Micron and tap into the global AI market.

D-Matrix, a new AI chip startup, challenges Nvidia’s dominance with its “Corsair” inference chip. Offering tenfold speed and fivefold energy savings for smaller AI tasks, Corsair utilizes a novel memory architecture integrating SRAM. While excelling in low-latency inference, its SRAM capacity limits handling massive models. D-Matrix targets interactive AI applications and partners with industry leaders for data center solutions, positioning itself as a significant player in the trillion-dollar AI inference market.

Broadcom’s Q2 earnings revealed revenue slightly below expectations and no upward revision to its AI chip sales target, causing a 15% stock drop. Despite strong year-over-year growth driven by custom AI chip demand, investors are scrutinizing the company’s ambitious $100 billion AI sales forecast. Broadcom clarified its focus on “chips only” and anticipates AI revenue tripling this quarter, but warned of staggered deliveries. The infrastructure software segment also saw mixed results.

Nvidia’s stock faces a critical test after a strong earnings report, as it pulls back to a key support level at $210-$215. Technical analysis suggests this zone, previously resistance, should now act as support, but a breach could re-establish it as resistance. Industry veteran Jim Cramer suggests Nvidia consider Apple’s capital return strategy to boost shareholder value and potentially reallocate investor positions given current market dynamics, despite strong long-term fundamentals.

Market participants are focused on strategic partnerships, corporate plans, and regulatory scrutiny. Key developments include intensifying AI chip competition with Microsoft and Anthropic in talks, Stellantis accelerating EV ambitions with a $70 billion investment plan, legislative gridlock nearing a June 1 deadline over DHS funding, a heated regulatory battle over prediction markets involving the CFTC and state authorities, and Oura, the smart ring maker, filing for a potential IPO.

Microsoft is reportedly in advanced talks to supply its AI chips to Anthropic, a major AI research firm. This potential deal would bolster Microsoft’s position in the competitive AI hardware market and help Anthropic address its significant compute needs. Anthropic is also diversifying its hardware strategy, with existing partnerships with Amazon and Google.

Nvidia CEO Jensen Huang admitted the company has “largely conceded” the Chinese AI chip market to Huawei due to US export restrictions. Despite this, Nvidia reported record financial results, with revenue soaring to $81.62 billion. Huang expressed no expectation of near-term market re-entry into China, focusing instead on global expansion and supporting its supply chain amidst massive AI economy growth.

Nvidia CEO Jensen Huang faces pressure on his earnings call to address Amazon and Alphabet’s growing custom AI chip development. Analysts urge him to adopt an offensive strategy, highlighting strong demand for Nvidia chips from a broad customer base, even from these hyperscalers. While acknowledging these competitors’ progress, Huang should project confidence in Nvidia’s market leadership and continued innovation to navigate the evolving AI hardware landscape.

Cerebras Systems’ IPO highlights strong demand for AI chips beyond Nvidia. The company, known for its large, custom-designed chips for AI inference, debuted with a valuation near $100 billion. While Nvidia dominates training, Cerebras targets the growing inference market with its Application-Specific Integrated Circuits (ASICs). This positions them against tech giants developing their own chips and other specialized startups like Groq and SambaNova. Cerebras now offers its solutions as a cloud service and has secured major deals with OpenAI and AWS.