Smartotics Investment Daily - 2026-07-07

📈 Market Overview

The technology investment landscape today is defined by a fascinating paradox: while hyperscalers and financial institutions publicly question the sustainability of AI spending, the semiconductor sector continues to deliver unprecedented financial results. Samsung Electronics reported a staggering 19x profit surge in Q2 2026, briefly overtaking NVIDIA as the world’s most profitable company—a testament to the AI-driven memory boom that shows no signs of abating. Meanwhile, SK Hynix’s impending US IPO is drawing aggressive positioning from top AI funds, with UBS projecting $15 billion in long-term capital inflows. The cognitive dissonance between bearish headlines from banks and hyperscalers—who themselves are the primary beneficiaries of AI infrastructure buildout—and the actual capital flows into AI and semiconductor assets suggests we are witnessing a structural shift rather than speculative froth. In China, AR smart glasses maker Even Realities closed a Pre-B round, signaling continued investor appetite for AI-enabled wearable hardware. The market is clearly bifurcating between companies with tangible AI revenue streams and those riding narrative alone.

💰 Funding Radar

1. Even Realities (逸文科技) - Pre-B Round (Undisclosed Amount)

Source: 36Kr

Deal Details:

Why It Matters:

The AR smart glasses market is experiencing a renaissance driven by two converging trends: the miniaturization of silicon photonics and the integration of on-device AI inference. Even Realities sits at the intersection of these trends. Unlike Meta’s Ray-Ban Stories, which rely on smartphone tethering for compute, Even Realities has developed proprietary waveguide optics and a custom ASIC for local LLM inference. This enables sub-100ms response times for AI queries without cloud dependency—a critical differentiator for enterprise adoption where data privacy is paramount.

The Chinese AR market is particularly interesting because of government support for “new quality productive forces” and the domestic supply chain for microLED displays and diffractive waveguides. Even Realities sources its display modules from BOE Technology and its AI chips from Horizon Robotics, creating a vertically integrated ecosystem that is less exposed to US export controls than competitors using Qualcomm or Broadcom components.

Competitive Positioning:

My Take:

Investment Thesis: Even Realities represents a compelling bet on the “AI glasses as the next smartphone” thesis. The company’s Pre-B round valuation is likely in the $200-300 million range based on comparable Chinese hardware startups. The key catalyst is the launch of their second-generation product in Q1 2027, which will feature a custom 3nm AI chip from Horizon Robotics, enabling real-time visual search and object recognition. If they can demonstrate 500,000+ unit annual shipments by 2028, a path to $1 billion revenue exists at an average selling price of $400.

Risk Factors:

  1. Competitive intensity: Apple, Meta, Google, and Samsung are all investing heavily in AR glasses. Even Realities lacks the distribution, brand, and R&D budget of these giants.
  2. Supply chain concentration: Dependence on BOE for displays and Horizon for chips creates single-point-of-failure risks.
  3. Regulatory uncertainty: China’s tightening control over AI applications and data localization could limit international expansion.
  4. Consumer adoption: AR glasses remain a niche category. Even Realities must prove that enterprise demand is sufficient to sustain growth before consumer adoption materializes.

Growth Potential: The global AR glasses market is projected to grow from $3.2 billion in 2025 to $28.6 billion by 2030 (Grand View Research). Even Realities, with its focus on AI-native hardware and enterprise use cases, could capture 3-5% of this market by 2030, implying $850 million to $1.4 billion in revenue. At a 5x revenue multiple (conservative for hardware), that suggests a $4-7 billion valuation potential—a 20-35x return from current levels if execution is flawless.

🏢 IPO & M&A Watch

SK Hynix US IPO: The AI Memory Giant Goes Public

Source: Wall Street CN

Deal Details:

Why It Matters:

SK Hynix is not just a memory manufacturer—it is the dominant supplier of HBM3E memory used in NVIDIA’s H200 and B200 AI accelerators. With an estimated 65% market share in HBM (High Bandwidth Memory), SK Hynix is arguably the most strategically important semiconductor company in the AI supply chain after NVIDIA and TSMC. The US IPO represents a landmark event for several reasons:

  1. AI Infrastructure Monetization: SK Hynix’s revenue from HBM grew 340% year-over-year in Q1 2026, accounting for 45% of total revenue. The IPO allows investors to gain pure-play exposure to AI memory, which is structurally different from traditional DRAM cycles.

  2. Geopolitical Arbitrage: By listing in the US, SK Hynix gains access to American institutional capital while maintaining its Korean manufacturing base. This is particularly important given the CHIPS Act’s incentives for domestic memory production—SK Hynix is building an advanced packaging facility in Indiana specifically for HBM.

  3. Valuation Benchmark: Current private market transactions value SK Hynix at approximately 15x forward earnings, compared to Samsung’s 18x and Micron’s 22x. The IPO could close this gap, especially if the company demonstrates continued HBM market share gains.

UBS’s $15 Billion Thesis: UBS analysts argue that SK Hynix is uniquely positioned to benefit from three structural trends:

My Take:

This is the most important tech IPO of 2026. SK Hynix’s US listing will serve as a barometer for investor sentiment on AI infrastructure spending. If the IPO prices at the top of the range and trades up, it validates the thesis that AI capital expenditure is not a bubble but a multi-year structural shift. If it disappoints, it will amplify concerns about peak AI spending.

Key Metrics to Watch:

Risk Factors:

Samsung Electronics: The Profit Surge That Changes Everything

Source: Wall Street CN

Financial Highlights:

Why This Matters for AI Investors:

Samsung’s profit surge is the clearest signal yet that the AI boom is not just about NVIDIA and hyperscalers—it is permeating the entire semiconductor ecosystem. Key implications:

  1. Memory is the new bottleneck: Samsung’s memory division alone generated more profit than most AI software companies. This validates the thesis that AI inference and training require massive memory bandwidth, creating a structural demand shift that decouples memory from its historical cyclicality.

  2. Foundry competition intensifies: Samsung’s 3nm GAA (Gate-All-Around) process is now yielding 70% of TSMC’s 3nm yields, a significant improvement from 50% in 2025. This positions Samsung as a viable second source for AI chips, particularly for cost-sensitive customers like Google and Amazon.

  3. AI smartphone monetization: Samsung’s mobile division profit growth was driven by Galaxy S26’s on-device AI features (real-time translation, AI photo editing, generative wallpapers). This demonstrates that AI can drive hardware upgrade cycles—a thesis that Apple will test with iPhone 17 in September.

My Take:

Samsung’s profit surge is sustainable, but the “world’s most profitable company” title is temporary. NVIDIA’s Q3 guidance (expected in August) will likely show net income of $12-13 billion, reclaiming the top spot. However, the key insight is that Samsung’s memory division now has structural pricing power that it lacked in previous cycles.

Investment Implications:

📊 Sector Analysis

🔥 Hot Sectors

1. High Bandwidth Memory (HBM)

The hottest subsector in semiconductors. SK Hynix’s US IPO and Samsung’s profit surge confirm that HBM is the most capacity-constrained and pricing-powerful segment of the AI supply chain. Key data points:

Investment vehicles: SK Hynix (IPO pending), Samsung Electronics (005930:KRX), Micron Technology (MU:NASDAQ)

2. AI Inference Hardware

As AI moves from training to inference, the hardware landscape is shifting:

Key metric: Inference-as-a-service revenue at AWS, Azure, and GCP grew 180% YoY in Q2 2026.

3. On-Device AI

Smartphones, PCs, and wearables with local AI processing:

Market opportunity: 1.2 billion AI-capable smartphones shipped in 2026 (IDC estimate).

❄️ Cooling Sectors

1. Autonomous Vehicle L4/L5

While robotaxi deployments continue (Waymo, Baidu Apollo), investment has cooled:

2. General-Purpose Robotics

Industrial robotics remains strong, but humanoid robotics funding has slowed:

🌟 Emerging Themes

1. AI Memory as an Asset Class

The concept of “memory-as-a-service” is gaining traction:

2. Silicon Photonics for AI Interconnects

As GPU clusters scale to 100,000+ units, electrical interconnects become a bottleneck:

Market size: $4.5 billion in 2026, growing to $15 billion by 2030 (LightCounting)

3. AI Chiplet Architectures

The shift from monolithic dies to chiplets is accelerating:

Key metric: Chiplet-based AI accelerators will account for 60% of AI chip revenue by 2028 (Yole Group).

🎯 Smartotics Portfolio Watch

NVIDIA Corporation (NVDA)

Current Status: Under pressure from AI bubble narrative Key Developments:

Our View: NVIDIA remains the most important AI company. The bubble narrative is overblown—NVIDIA’s data center revenue is backed by hyperscaler CapEx commitments totaling $250 billion through 2028. However, we are reducing position size by 5% to take profits and rebalance into SK Hynix IPO.

Samsung Electronics (005930:KRX)

Current Status: Strong buy after Q2 earnings beat Key Developments:

Our View: Samsung offers the best risk/reward in semiconductors. At 12x forward earnings with 35% EPS growth, it trades at a discount to peers. We are increasing position by 10%.

SK Hynix (Pre-IPO)

Current Status: Building position ahead of US listing Our View: We recommend a 3% portfolio allocation to SK Hynix at IPO. The company’s HBM dominance, US manufacturing expansion, and structural pricing power make it a core AI holding. Target price: 30% premium to IPO price within 12 months.

🔮 Next Week Preview

July 13-17, 2026: Key Events

  1. TSMC Q2 2026 Earnings (July 13)

    • Expected revenue: NT$750 billion ($23.5 billion), up 40% YoY
    • Key metric: 3nm revenue contribution (expected 25% of total)
    • Guidance for Q3: Will set the tone for AI semiconductor demand

  2. World AI Conference Shanghai (July 14-16)

    • Keynote speakers: Jensen Huang (NVIDIA), Sam Altman (OpenAI), Robin Li (Baidu)
    • Expected announcements: New Chinese AI chips, LLM releases, robotics demos
    • Watch for: Huawei’s Ascend 920 AI GPU launch

  3. AMD Data Center AI Summit (July 15)

    • Focus: MI400 architecture, ROCm software ecosystem
    • Key metric: MI300X vs. NVIDIA H200 benchmark comparisons
    • Expected: Customer wins (Microsoft, Oracle, Meta)

  4. U.S. CHIPS Act Funding Announcements (July 17)

    • Expected: $5 billion in grants for advanced packaging facilities
    • Beneficiaries: Intel (Ohio), Samsung (Texas), TSMC (Arizona)
    • Impact: Domestic semiconductor supply chain acceleration

Investor Action Items:

📝 Final Thoughts

The AI investment landscape in July 2026 is characterized by a fascinating tension: public skepticism from banks and hyperscalers versus record-breaking financial results from semiconductor companies. The “AI bubble” narrative, while attention-grabbing, ignores the fundamental reality that AI infrastructure spending is backed by concrete ROI—hyperscalers are seeing 3-5x returns on AI compute investments through cloud revenue growth and internal productivity gains.

The SK Hynix IPO will be the defining event of H2 2026. If successful, it will unlock a wave of semiconductor IPOs (including potential offerings from Arm, ASML, and Tokyo Electron spin-offs). If it falters, it will validate the bubble narrative and trigger a 15-20% correction in AI semiconductor stocks.

Our recommendation: Stay long AI infrastructure, but rotate from pure-play GPU exposure (NVIDIA) to diversified semiconductor plays (Samsung, SK Hynix) and emerging AI hardware themes (silicon photonics, chiplet architectures). The next leg of the AI bull market will be driven by inference at scale, not training—and that requires memory, interconnects, and packaging as much as it requires GPUs.

Smartotics Portfolio Allocation:

This report is for informational purposes only and does not constitute investment advice. Past performance is not indicative of future results.

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.