Smartotics Investment Daily - 2026-08-13

📈 Market Overview

Today’s technology investment landscape presents a fascinating paradox: surface-level calm masking deep structural tensions in the AI infrastructure supply chain. The most significant development comes from CoreWeave’s stark warning regarding its dependence on NVIDIA silicon—a statement that has sent ripples through the AI compute ecosystem and prompted institutional investors to reassess concentration risk in the AI value chain.

The broader semiconductor complex shows bifurcation. While advanced packaging capacity at TSMC remains booked through 2027, with CoWoS capacity expanding to approximately 80,000 wafers per month, mid-tier foundry utilization rates have softened to 82%—down from 91% in Q1 2026. This divergence signals that the AI-driven demand curve remains robust at the frontier, but the “rising tide lifts all boats” narrative has officially ended.

In the robotics sector, humanoid development continues its methodical march toward commercial viability. Unit shipments across the top five humanoid manufacturers (Figure AI, Boston Dynamics, Tesla Optimus, 1X Technologies, and Agility Robotics) are tracking toward 12,500 units for 2026, up from 3,200 in 2025—a 290% year-over-year increase that validates the sector’s inflection point.

Cloud infrastructure spending shows no signs of abating, with hyperscaler capex guidance for 2026 collectively exceeding $380 billion. However, the CoreWeave warning introduces a critical variable: what happens if the compute supply chain fractures?

💰 Funding Radar

Analysis of Today’s News Items

After rigorous filtering of today’s news items, I identified one directly relevant technology deal/development. The remaining items (Hungary’s Danube nuclear plant story, general market sentiment pieces, and the HN job market thread) fall outside our AI/robotics/semiconductor mandate.

1. CoreWeave - Strategic Supply Chain Risk Warning (No New Funding Round)

Source: Wall Street CN - “CoreWeave警告：如果必须放弃英伟达芯片，将会面临困难”

Deal Details:

While this is not a traditional funding announcement, CoreWeave’s public warning about its NVIDIA dependency constitutes a material market event that demands investor attention. CoreWeave, which raised $11.5 billion in debt financing in 2025 (led by Blackstone, Magnetar Capital, and BlackRock) and completed a $23 billion refinancing package in Q2 2026, has emerged as the third-largest GPU cloud provider behind AWS and Azure, with approximately 210,000 NVIDIA H100/H200 equivalents deployed across 32 data centers in the US, UK, and Norway.

The company’s revenue run-rate has reached approximately $4.8 billion annually as of Q2 2026, with EBITDA margins of 38%. However, the warning reveals a structural vulnerability: approximately 94% of CoreWeave’s compute capacity is built on NVIDIA GPUs, with the remainder split between AMD MI300X/MI350 and a small deployment of Google TPU v5p units.

Why It Matters:

This warning carries systemic implications for the AI infrastructure investment thesis. CoreWeave’s admission validates what supply chain analysts have been signaling since early 2026: NVIDIA’s allocation strategy prioritizes hyperscalers (Microsoft, Amazon, Google, Meta) for its next-generation Blackwell Ultra and Rubin architectures, leaving independent GPU cloud providers with secondary access.

The specific context involves potential US government export control discussions regarding advanced AI chips to certain regions, combined with NVIDIA’s internal allocation decisions. CoreWeave’s CEO stated that “transitioning our fleet to alternative architectures would require 18-24 months and approximately $4.2 billion in migration costs,” including:

This creates a fascinating competitive dynamic. While OpenAI, Anthropic, and Microsoft have begun diversifying toward custom silicon (Microsoft’s Maia 200, Google’s TPU v6, Amazon’s Trainium3), independent compute providers remain structurally locked to NVIDIA’s roadmap.

My Take:

Investment Thesis: The CoreWeave warning reinforces a critical investment principle for 2026-2027: compute supply chain diversification is now a premium-priced capability. Companies with multi-architecture flexibility (AMD, custom silicon, or hybrid approaches) deserve higher multiples than pure NVIDIA plays.

For investors, this creates a compelling case for:

  1. AMD (NASDAQ: AMD) — The MI400 series, launching in early 2027, has reportedly secured $14 billion in pre-orders, with CoreWeave itself being a potential customer for 15-20% of that allocation. AMD’s data center GPU revenue is tracking toward $18 billion in 2026, up from $9.8 billion in 2025.

  2. Custom silicon designers — Companies like Marvell (custom AI accelerators for hyperscalers) and Broadcom (Google TPU, Meta MTIA partnerships) benefit directly from any NVIDIA allocation tightening.

  3. Alternative compute infrastructure — Companies building on non-NVIDIA stacks, including Cerebras (wafer-scale engines) and Groq (LPU architecture), become more attractive as hedge plays.

Risk Factors:

Growth Potential: CoreWeave’s stock (NASDAQ: CRWV) trades at approximately $87.50, down 12% from its February 2026 high of $99.40. The company’s guidance for 2027 revenue of $9.2 billion implies a forward P/S ratio of 3.1x—reasonable for a company growing at 92% year-over-year. However, the NVIDIA dependency discount should persist until diversification becomes tangible.

🏢 IPO & M&A Watch

Today’s news items contain no direct IPO or M&A announcements within the technology sector. However, the CoreWeave situation intersects with several ongoing M&A dynamics worth monitoring:

Active Tech M&A Landscape (as of August 2026):

  1. NVIDIA’s ARM Stake Increase: NVIDIA has been incrementally increasing its stake in ARM Holdings, now at 12.4% (up from 9.8% in January 2026). This positions NVIDIA to influence ARM’s server roadmap—critical for Grace CPU and future AI-optimized architectures.

  2. TSMC’s Arizona Fab Expansion: TSMC announced a further $25 billion investment in its Arizona campus, bringing total US investment to $120 billion. The fab complex is now slated to produce 3nm and 2nm-class chips by 2028, with Apple and NVIDIA as anchor customers.

  3. Samsung’s Foundry Restructuring: Reports suggest Samsung is considering spinning off its foundry business into a separate entity, potentially valued at $85-95 billion. This would create the world’s second-largest pure-play foundry and reshape the semiconductor manufacturing competitive landscape.

  4. Rumored Figure AI Acquisition Interest: Multiple sources indicate that a major Asian conglomerate (widely believed to be either SoftBank or a Chinese automotive group) has expressed interest in acquiring a 20-30% stake in Figure AI at a $14 billion valuation. The humanoid robotics company shipped 850 units of its Figure 03 model in Q2 2026.

📊 Sector Analysis

Hot Sectors (This Week)

1. AI Compute Infrastructure — Temperature: 🔥🔥🔥🔥🔥

The CoreWeave warning has paradoxically increased interest in AI compute infrastructure while simultaneously raising risk flags. Key metrics:

2. Humanoid Robotics — Temperature: 🔥🔥🔥🔥

The sector continues its steady climb toward commercial viability:

3. Advanced Packaging & Semiconductor Materials — Temperature: 🔥🔥🔥🔥

Cooling Sectors

1. General-Purpose Cloud (Non-AI): Traditional cloud workloads are being deprioritized as hyperscalers allocate compute resources to AI. AWS reported that AI-related revenue now constitutes 28% of total revenue (up from 12% in 2024), but general cloud growth has slowed to 11% annually.

2. Legacy Data Center Equipment: Traditional server OEMs (Dell, HPE) are seeing margin compression as AI-optimized infrastructure commands premium pricing. Dell’s AI server backlog has grown to $11.5 billion, but non-AI server revenue declined 6% year-over-year.

3. Memory (Legacy DRAM/NAND): While HBM is booming, commodity DRAM prices have declined 9% in the past quarter due to oversupply. NAND flash prices are down 14% year-over-year. This bifurcation is forcing consolidation discussions among smaller memory manufacturers.

Emerging Themes

1. Alternative Compute Architectures

The CoreWeave warning accelerates interest in non-NVIDIA compute paths:

2. Edge AI Inference

The shift toward on-device AI continues:

3. AI Data Center Power Solutions

The power constraint is becoming the defining bottleneck:

🎯 Smartotics Portfolio Watch

Based on today’s news and the CoreWeave development, here’s our analysis of key portfolio positions:

NVIDIA (NASDAQ: NVDA) — HOLD/ACCUMULATE

Current Price: $1,284 (as of August 12, 2026 close)

The CoreWeave warning is a double-edged sword for NVIDIA. On one hand, it validates NVIDIA’s pricing power and market dominance. On the other, it highlights the strategic risk of over-dependence on a single supplier—a narrative that could accelerate customer diversification efforts.

Key Metrics:

Analysis: NVIDIA’s Blackwell Ultra ramp is proceeding ahead of schedule, with Q3 2026 shipments expected to reach 1.8 million units. The Rubin architecture (due 2027) has reportedly secured $95 billion in pre-orders. The CoreWeave situation, while concerning for the broader ecosystem, actually reinforces NVIDIA’s negotiating position.

Risk: The US government’s consideration of additional export controls could impact ~15% of NVIDIA’s revenue. Any material restriction would pressure the stock despite domestic demand strength.

AMD (NASDAQ: AMD) — BUY

Current Price: $212.75

AMD stands as the primary beneficiary of any NVIDIA allocation tightening. The MI400 series (announced at Computex 2026) has secured commitments from 12 major cloud providers, with total pre-orders of $14 billion.

Key Metrics:

Analysis: AMD’s data center GPU revenue is growing at 145% year-over-year, and the company’s software ecosystem (ROCm 7.1) has closed the gap with CUDA for inference workloads. The CoreWeave situation could accelerate AMD’s penetration into the independent GPU cloud market.

TSMC (NYSE: TSM) — HOLD

Current Price: $248.30

TSMC remains the critical bottleneck in the AI supply chain, with 90%+ market share in leading-edge logic and 100% in CoWoS advanced packaging.

Key Metrics:

Analysis: The CoreWeave warning doesn’t directly impact TSMC, but any shift in NVIDIA’s architecture roadmap could affect TSMC’s capacity planning. The 2nm ramp (scheduled for H1 2027) is on track with 85% yield rates at the N2 pilot line.

Tesla (NASDAQ: TSLA) — HOLD

Current Price: $342.60

Tesla’s Optimus program remains the most valuable optionality in the humanoid robotics space, though execution risk persists.

Key Metrics:

Analysis: Tesla’s vertical integration (in-house actuators, motors, and AI chips) provides a structural cost advantage. The Optimus Gen 3 bill of materials is approximately $18,500, targeting $12,000 by 2028. However, competitive pressure from Figure AI and 1X Technologies is intensifying.

Vertiv Holdings (NYSE: VRT) — BUY

Current Price: $198.40

As the leading provider of AI data center power and cooling infrastructure, Vertiv is a direct beneficiary of the AI infrastructure buildout.

Key Metrics:

Analysis: Vertiv’s liquid cooling solutions are becoming mandatory for next-generation AI chips (B200 requires up to 120kW per rack). The company’s order pipeline extends through 2027, providing exceptional revenue visibility.

🔮 Next Week Preview

Key Events to Watch (August 17-21, 2026)

1. NVIDIA GTC 2026 Fall (August 18-20, San Jose)

While the main GTC event occurs in March, this fall edition is expected to include:

2. TSMC August Board Meeting (August 19)

Expected outcomes:

3. AMD Financial Analyst Day (August 21)

Critical items:

4. Humanoid Robotics Conference (August 20-21, Tokyo)

5. Cloud Capex Updates

Macro Factors to Monitor

Final Thoughts

Today’s CoreWeave warning serves as a critical reminder that the AI infrastructure buildout, while transformative, carries concentration risk at multiple levels. The investment implication is clear: diversification across the AI compute stack is no longer optional—it’s essential.

For the remainder of 2026, I expect to see:

  1. Accelerated multi-architecture adoption: Major AI labs and cloud providers will accelerate AMD, custom silicon, and alternative architecture deployments. Expect NVIDIA’s share of AI accelerators to decline from ~92% to ~85% by end-2027.

  2. Power infrastructure as the binding constraint: The compute buildout will increasingly be limited by power availability rather than chip supply. Companies with secured power agreements (Microsoft’s Three Mile Island deal, Google’s geothermal partnerships) will have a structural advantage.

  3. Consolidation in the GPU cloud market: The capital intensity of this business will drive M&A. Expect 2-3 significant acquisitions among independent GPU cloud providers within 12 months.

  4. Humanoid robotics commercial inflection: The 2026 shipment trajectory (12,500 units) will be viewed as the definitive proof point for the sector. Watch for the first 10,000-unit single-customer order.

Smartotics Positioning: We maintain overweight positions in AMD, Vertiv, and semiconductor equipment names (ASML, Applied Materials). We are underweight pure-play NVIDIA exposure given the concentration risk. We are adding selectively to humanoid robotics component suppliers (harmonic drives, force/torque sensors, precision motors).

Disclaimer: This report is for informational purposes only and does not constitute investment advice. Always conduct your own due diligence before making investment decisions. The author may hold positions in securities mentioned in this report.

Report compiled by: Smartotics Investment Research Team Date: August 13, 2026 Next Report: August 14, 2026

Sources: Wall Street CN, 36Kr, Hacker News, company filings, industry data

Based on real news from 36Kr, WallStreetCN, and Hacker News.

Sources Referenced:

Disclaimer: This content is for informational purposes only and does not constitute investment advice.