Highlighted Companies
KEY COMPANIESAnthropic filed a confidential S-1 with the SEC on June 1 — stunning the market by moving before OpenAI, which had been widely assumed to hold the first-mover IPO advantage. The filing comes just days after Anthropic closed its $65B Series H at a $965B post-money valuation. The confidential filing buys Anthropic time to market a cleaner growth story while the SEC reviews it: $47B annualized revenue run-rate, new “Mythos” model in weeks, and enterprise momentum from Claude integrations with Salesforce Agentforce and Google Gemini Enterprise. The IPO race is now live — OpenAI is expected to file its own S-1 in H2 2026 at approximately $1T. Whoever lands first sets the valuation anchor for the entire AI investment category.
Read More →Salesforce and Google Cloud activated their expanded AI partnership in June, with Agentforce natively integrating Gemini 3.5 Flash via the Atlas Reasoning Engine — giving Salesforce agents multimodal capability (text, image, video) and access to years of customer history. Agents can now execute end-to-end workflows across Salesforce, Slack, and Google Workspace without handoffs or context loss. The A2A (Agent-to-Agent) protocol governing cross-platform routing is now in production at 150+ organizations. This is a direct response to Microsoft Copilot’s distribution advantage: Salesforce is betting that deep platform integration with Google’s model stack will close the gap on enterprise agentic workflows before Microsoft’s ecosystem becomes entrenched.
Read More →Microsoft Build 2026 introduced “Autopilots” — a new category of always-on AI agents that operate autonomously under governed Entra ID identities, acting on behalf of users across Microsoft 365, Teams, and Azure. Foundry Agent Service reached General Availability. Simultaneously, OpenAI’s GPT-5.5 went GA in Microsoft Foundry on June 3 — featuring deeper long-context reasoning, improved agentic execution accuracy, and greater token efficiency, with GPT-5.5 Pro as the premium tier. The effect: Microsoft is not building an AI assistant product — it is building an AI operating system. Autopilots with Entra ID governance turn the enterprise identity layer into the control plane for autonomous AI, creating a Microsoft moat that is structurally difficult for standalone AI vendors to replicate.
Read More →AI & Semiconductors
AI & CHIPSNVIDIA unveiled the RTX Spark Superchip at Computex 2026 — a 70-billion-transistor part on TSMC 3nm that fuses an Arm CPU (up to 20 cores) and a Blackwell GPU (6,144 CUDA cores) with 128GB of LPDDR5X unified memory, 300 GB/s bandwidth, and 1 petaflop of AI performance. RTX Spark can run 120B-parameter local LLMs with million-token context windows — effectively turning a consumer laptop into a private inference cluster. Dell, HP, Lenovo, Asus, MSI, and Microsoft Surface devices launch this fall. The strategic read: Jensen Huang is not building a PC chip — he is establishing a third compute tier between the cloud and the edge, purpose-built for agentic workloads that require local context, low latency, and data sovereignty.
Read More →Broadcom’s fiscal Q2 2026 earnings beat on revenue ($22.19B vs. $22.13B expected) and AI semiconductor revenue grew 143% YoY to $10.80B — but Q3 AI chip guidance of $16B came in $1.2B below the $17.2B analyst consensus, and the company declined to raise its full-year AI outlook. The market read this as peak cycle signaling: Broadcom stock fell 14% on June 4, dragging AMD, Intel, and Micron down with it. The Nasdaq dropped 4.18% for its worst session since April 2025. The read: this is not a demand problem — it is a visibility problem. Broadcom’s custom AI chip business (Google TPUs, Meta MTIA) is inherently lumpy quarter-to-quarter. The selloff is an overreaction to a guidance cadence issue, not evidence of AI infrastructure spending deceleration.
Read More →AMD’s Venice server processor — the world’s first high-performance computing chip to enter mass production on a 2nm process — is now in full ramp-up at TSMC foundries. Venice succeeds Turin (3nm) and marks a full generation of node advancement in one product cycle. AMD also committed $10B+ to Taiwan’s AI semiconductor ecosystem at Computex, including capacity reservations and OSAT partnerships. The 2nm milestone matters because it confirms that TSMC’s N2 node is commercially viable at HPC-class die sizes, expanding its applicability beyond mobile. For AMD, it narrows the manufacturing gap with NVIDIA’s Blackwell lineup and positions EPYC Venice as the performance leader in AI inference workloads that prize CPU-side memory bandwidth.
Read More →Markets & Tech Stocks
S&P 500 · NASDAQ · MARKETSThe Nasdaq dropped 4.18% to 25,709 on June 5 — its worst single-session decline in over a year — as Broadcom’s disappointing AI chip guidance combined with a stronger-than-expected May jobs report to unwind two compounding trades simultaneously. The jobs data pushed Treasury yields higher (reducing the rate-cut probability already priced into tech valuations), while Broadcom’s miss created a “sell the sector” response in semiconductors. The S&P 500 fell 2.64% to 7,383. The Dow outperformed on defensive rotation, with Consumer Staples up over 2%. The selloff is best read as a valuation reset after AI earnings season peaked: the sector had run to new all-time highs on May 26 — the drawdown reflects profit-taking on a crowded trade, not a fundamental shift in AI spending.
Read More →Alphabet executed an $84.75 billion equity capital raise on June 1 — comprising $15B in mandatory convertible preferred, $15B in common stock, a $40B ATM program, and a $10B Berkshire Hathaway private placement. The declared purpose: fund AI compute infrastructure to meet “unprecedented customer demand.” The raise resets the benchmark for how much capital hyperscalers are willing to dilute shareholders to win the AI infrastructure race. Notably, Berkshire’s participation — Warren Buffett’s firm historically avoiding tech — signals that AI infrastructure is now considered value-category infrastructure, not speculative technology. Meta is now weighing a similar equity raise of tens of billions to fund its own $125-145B capex plan.
Read More →The S&P 500 closed above 7,600 for the first time on June 2, driven by continued semiconductor strength and the Alphabet capital raise signaling hyperscaler AI conviction. The milestone came at the apex of the post-May-26 rally that had already set new all-time highs on Micron’s trillion-dollar market cap day. Three days later, the Broadcom guidance miss erased roughly 2.6% from the index — leaving the S&P at 7,383. The pattern is textbook AI earnings-cycle mechanics: a build into the peak print (NVIDIA Q1 FY2027), followed by broadening relief rallies on secondary plays (Snowflake, Micron), capped by a guidance-miss catalyst that forces near-term repricing without changing the structural bull thesis. NVIDIA’s CFO confirmed at the BofA Tech Conference that addressable markets can double with new chip generations.
Read More →Supply Chain & Commodities
CHIPS · MATERIALS · FREIGHTCombined hyperscaler capex (Microsoft, Alphabet, Meta, Amazon) is tracking toward $725 billion in 2026 — with approximately 75% earmarked for AI infrastructure. Amazon alone expects $200B in capex this year; Meta is at $115-145B. The supply constraint that has replaced GPU availability as the binding limit: power. Meta’s nuclear PPA, Amazon’s nuclear offtake expansion, and Microsoft’s Three Mile Island restart confirm that nuclear is no longer aspirational — it is being contracted at scale. Utility-scale AI data centers are now incompatible with conventional grid infrastructure in most U.S. markets, making nuclear, geothermal, and dedicated gas-fired power the de facto energy supply chain for the next generation of compute clusters.
Read More →Meta Platforms is reportedly weighing a multi-billion dollar equity offering — potentially in the tens of billions — to fund a capex budget that has expanded to $125-145B for 2026, according to the Financial Times. The potential offering follows Alphabet’s $84.75B capital raise by less than a week. The structural implication: hyperscalers are no longer funding AI infrastructure purely from operating cash flows — they are going to public markets to accelerate spend beyond what organic generation allows. This shifts the AI infrastructure financing model from internal capital allocation to a capital markets competition, with each hyperscaler effectively raising equity to outpace rivals in GPU and data center buildout. Meta has not confirmed the offering and called the reporting speculative.
Read More →NVIDIA CFO Colette Kress told the Bank of America Global Technology Conference that the company’s addressable market can double as successive GPU generations come to market — citing Blackwell’s continued data center dominance, the RTX Spark platform’s entry into the PC and personal AI market, and the emerging physical AI (robotics, autonomous systems) segment as three distinct expansion vectors. BofA reiterated its Buy rating and $350 price target. The market sizing argument: if data center AI is already a $75B quarterly revenue segment at NVIDIA and Spark opens a new consumer/enterprise endpoint market at scale, the total addressable compute spend that flows through NVIDIA silicon meaningfully exceeds even the current consensus models — particularly as sovereign AI programs and ACIE (non-hyperscaler cloud) buyers continue to scale.
Read More →