☀️ Morning Briefing
Highlighted Companies
KEY COMPANIESGoogle consolidated its entire AI stack at Cloud Next 2026, unifying Vertex AI and Agentspace into a single Gemini Enterprise product. The Agent2Agent (A2A) protocol reached v1.2 with cryptographic agent verification, and Salesforce, SAP, ServiceNow, and Workday are all running A2A in production — a significant adoption signal for Google’s bid to be the enterprise AI control plane.
Read More →Salesforce was named as a production partner for Google’s A2A protocol at Cloud Next 2026, with Gemini embedded more deeply across Agentforce workflows. The move strengthens Salesforce’s multi-cloud AI narrative heading into Q1 FY27 earnings season and reflects the broader pattern of enterprise SaaS vendors layering external AI onto their platforms rather than building proprietary models.
Read More →ServiceNow posted Q1 subscription revenue of $3.671B (+19% YoY) and raised its full-year guidance midpoint to $15.755B — a clean beat with no asterisks. More strategically, the company announced AI, data, security, and governance are now built into every product by default, not a separate SKU. The “AI Control Tower” governance layer, launching at Knowledge 2026 next week, is positioned as the shared oversight plane for agents across Google Cloud, Oracle, and Visa partnerships.
Read More →The Wall Street Journal reported OpenAI missed multiple monthly revenue targets in early 2026, falling behind Google Gemini in consumer and Anthropic in enterprise/coding. Finance chief Sarah Friar warned internally that accelerating revenue is necessary to fund $600B in compute commitments through 2030. OpenAI called the report “prime clickbait,” but the market reaction — across chips, cloud infrastructure, and Japan’s SoftBank — revealed how tightly the AI capex thesis is tied to OpenAI monetization proof points.
Read More →AWS’s April announcements included Amazon Quick, a cross-platform agent designed to work natively across Slack, Teams, Outlook, Gmail, Salesforce, and ServiceNow — positioning it as a productivity-layer competitor to Microsoft Copilot. The product signals AWS’s intent to compete at the application layer rather than cede that ground entirely to Microsoft and Salesforce, though it arrives later than Copilot and Agentforce to market.
Read More →AI & Semiconductors
AI & CHIPSNVIDIA’s Rubin platform — now comprising seven chips including the new Groq 3 LPU — is in full production with cloud deployments beginning H2 2026. NVIDIA estimates Blackwell and Rubin combined will generate $1 trillion in chip sales across 2026–2027. CoreWeave, AWS, Google Cloud, and Microsoft are first movers. The demand picture remains robust despite OpenAI monetization noise, as inference workloads are diversifying well beyond ChatGPT.
Read More →Google’s two new AI processors announced at Cloud Next signal an accelerating hyperscaler push toward chip independence. The TPU 8t and 8i push inference and training performance deeper in-house, reducing NVIDIA reliance for Google’s own AI workloads and potentially for GCP customers. The pattern is consistent across the hyperscaler cohort: AWS (Trainium 3), Google (TPU 8), Microsoft (Maia 2), and Meta (MTIA) are all reducing external GPU dependency at scale.
Read More →Anthropic is exploring custom silicon to manage Claude compute costs and reduce NVIDIA dependency, mirroring the hyperscaler playbook. Simultaneously, Amazon announced an expanded chip deal with Anthropic under which the AI lab commits more than $100 billion in AWS technologies over the next decade. The pairing — build your own chips while deepening the primary cloud relationship — reflects the dual strategy of cost control and scale access that every frontier AI lab is navigating.
Read More →The WSJ’s report on OpenAI’s missed targets exposed how concentrated the AI chip demand thesis is in a small number of bellwether customers. NVIDIA, AMD, and Arm all sold off sharply, with Arm’s decline the steepest — reflecting its high valuation premium tied to the AI growth story. The episode is a useful stress test: chip demand from AWS, Google, Microsoft, and Meta remains structurally intact, but the market reprices on OpenAI signal because it’s the most visible AI monetization barometer.
Read More →The picks-and-shovels story is widening beyond NVIDIA. AMD, Broadcom, and Intel chips have all been formally selected for AI infrastructure deployments across OpenAI, Anthropic, Meta, Microsoft, and Amazon. NVIDIA retains a dominant share of accelerator demand, but the growing multi-supplier reality reflects both hyperscaler hedging and the maturing inference market where workload-specific chips are increasingly cost-competitive.
Read More →Markets & Tech Stocks
S&P 500 · NASDAQ · MARKETSThe S&P 500 closed at 7,209 (+1.02%) on April 29 and the Nasdaq at 24,892 (+0.89%), capping a historically strong month driven by a clean Q1 earnings season and easing geopolitical tensions. The S&P is +4% YTD and the Nasdaq +5% — solid performance against a backdrop of elevated AI capex uncertainty. The record close establishes a new base from which further multiple expansion will require either rate cuts or sustained AI monetization acceleration.
Read More →Apple’s blowout Q2 — $111.2B revenue vs. $109.7B expected, EPS $2.01 vs. $1.95 — was led by iPhone’s best March quarter ever ($58B, +22%) and Services at $30.98B, a record. China revenue surged 28% to $20.5B. Gross margin expanded to 49.3%. The $100B buyback authorization and the confirmed September 1 start for CEO John Ternus make this the cleanest handoff quarter in Apple’s recent history, removing a key overhang.
Read More →April 28 was the sharpest AI-related single-day selloff in months, cutting across chip stocks, cloud infrastructure plays, and Japan-listed AI investors. SoftBank’s -10% drop in Tokyo reflected its concentrated OpenAI exposure. Oracle’s -4% move underscored the revenue risk embedded in its $300B, five-year compute contract with OpenAI. The selloff reversed partially on April 29 as the broader market reclaimed record levels, suggesting investors separated OpenAI-specific risk from the structural AI capex case.
Read More →The April 28 session revealed a market repricing the risk premium differently across tech sub-sectors. Diversified hyperscalers (Apple, Alphabet, Microsoft) held up better than pure-play AI infrastructure names, while the Nasdaq’s underperformance relative to the Dow signals the market is beginning to distinguish between AI enablers with durable cash flows and companies whose valuations are predicated solely on AI monetization timelines that are proving harder to hit.
Read More →Markets are opening May with positive momentum off Apple’s earnings beat and oil price declines supporting the macro backdrop. The constructive open follows April’s record close and suggests the OpenAI selloff was digested as idiosyncratic rather than systemic. With the Fed meeting this week and Q1 earnings season nearly complete, May’s narrative will likely pivot to rate trajectory signals and early reads on enterprise software demand.
Read More →Supply Chain & Commodities
CHIPS · MATERIALS · FREIGHTGoogle, Amazon, Microsoft, and Meta collectively plan $725B in 2026 capex — a 77% increase over the prior record of $410B in 2025. Approximately 75% is earmarked for AI infrastructure. Amazon’s $200B commitment alone exceeds the total combined hyperscaler spend from three years ago. The supply chain implication is straightforward: demand for power, cooling, DRAM, HBM, and networking gear is structurally elevated well beyond what any single fab cycle can accommodate.
Read More →HBM3E is sold out through the end of 2026 with no relief anticipated before 2027. Memory prices surged 80–90% quarter-over-quarter as AI demand reallocated TSMC and SK Hynix capacity away from consumer DRAM toward high-margin HBM. The HBM market is projected to reach $100B by 2030 — a fourfold increase — driven by the requirement that HBM3E consumes roughly 3x the wafer supply per gigabyte compared to DDR5. Server and PC hardware pricing is expected to stay elevated for the next two to three quarters.
Read More →Microsoft’s disclosure of an $80B unfulfilled Azure order backlog is the starkest evidence yet that data center demand is structurally exceeding build capacity — and the bottleneck is power, not money. The company has set its 2026 capex at $190B, with $25B of that specifically attributable to rising memory and component costs. Amazon’s free cash flow is projected to turn negative this year on the same dynamic. The power permitting and grid interconnection timelines are now the binding constraint for AI infrastructure scaling.
Read More →Micron has publicly acknowledged it cannot meet demand from key customers, estimating it can fulfill roughly half to two-thirds of expected orders even while raising capex and planning new capacity. Its new US fab (ID1 in Idaho) won’t start production until 2027, providing no near-term relief. The IDC projects a global memory shortage crisis that will ripple through smartphone and PC markets alongside the AI server buildout, compressing supply across every price point and application segment simultaneously.
Read More →Geopolitical friction is creating regional pricing disparities for AI chips and components, with tariffs adding layers of cost and procurement complexity that chip brokers and procurement teams are struggling to model. Separately, TSMC’s CoWoS advanced packaging — the critical process that stacks HBM on GPUs — remains the most constrained node in the AI chip supply chain through at least 2027. No amount of fab capex resolves the packaging bottleneck without TSMC-specific capacity expansion, which is moving more slowly than chip die production.
Read More →