1. Humanoid Robot Mass Production Era Declared — Tesla Optimus Gen-3 & Unitree G1 Deployed in the Field, China Industrial Robot Exports +42% YoY
• Core Source
“Tesla’s Fremont factory has begun producing the Optimus Gen-3, and Unitree’s G1 has been deployed for actual operations at Tokyo’s Haneda Airport.”
“The market assesses that as mass production plans of global and Chinese leading companies are becoming concrete and China’s supply chain continues its technological breakthroughs, the industry has passed the technology verification stage and entered a full-scale commercial phase.”
“China’s total robot exports in Q1 reached approximately 11.32 billion yuan, with industrial robot exports in particular surging 42% year-over-year.”
“According to Wind data, the ‘China Humanoid Robot Index’ has risen approximately 19.5% since April 7.”
“Piper Sandler maintained its ‘Overweight’ rating and $500 price target on Tesla, implying that robot-related businesses can generate approximately $100 of value per share.”
• Expected Impact
2026 is likely to be recorded as a historic turning point where the humanoid robot industry stepped out of laboratories and began actual deployment on factory floors, logistics centers, and airport grounds. Tesla’s Optimus Gen-3 entered mass production at the Fremont factory, and Unitree’s G1 was put into actual ground maintenance operations at Haneda Airport. This is significant not merely as a technology demonstration, but because it marks the first time the bar of commercial operation has been cleared.
From an investor perspective, precisely identifying the beneficiary segments within the value chain is critical. The market analyzes that core component manufacturers such as reducers, servo motors, sensors, and precision parts are most likely to benefit first. Harmonic reducers, RV reducers, and tactile/torque/visual sensors that go into each robot have high technological barriers to entry and a limited number of capable suppliers, meaning that as production volumes surge, component supply bottlenecks may emerge first.
In China’s case, industrial robot exports already surged 42% year-over-year in Q1, with robots beginning to establish themselves as a new export driver. In particular, Shenzhen has formed an industrial cluster of 100,000 related companies, meaning China’s supply chain may possess both cost competitiveness and speed in the global robot mass production race.
Piper Sandler analyzed that at a Tesla share price of around $400, the Optimus business is essentially a free option. The robot business can generate approximately $100 of value per share, and even this figure may be conservative. Over the long term, Tesla’s Inference-as-a-Service strategy has the potential to be worth more than the sum of all its current automotive and energy businesses combined. The physical AI market, where AI expands into the physical world, is expected to fully emerge from 2028, and we are now at the very beginning of that trajectory.
2. TSMC Maximizes Pricing Power from Monopolistic AI Demand — NVIDIA Purchase Commitment Surging from $16B to $95B, Revenue Growth Forecast to Exceed 30%
• Core Source
“With big tech companies executing a total of $725 billion in CAPEX this year to secure AI chips, the shortage of cutting-edge semiconductors is deepening further. In particular, TSMC, which effectively monopolizes the most advanced processes, has secured strong pricing power.”
“TSMC’s Q1 gross margin rose to 66%, and the CEO expressed strong confidence that revenue growth this year will exceed 30%.”
“In particular, NVIDIA’s purchase commitment to TSMC is forecast to surge from $16 billion to approximately $95 billion within the next two years.”
“Some customers have reportedly been prepaying for years to secure advanced packaging and wafer production capacity.”
• Expected Impact
TSMC currently occupies a virtually irreplaceable position in the global AI chip supply chain. As major fabless companies including NVIDIA, Apple, and AMD all depend on TSMC’s most advanced processes, a situation of demand structurally exceeding supply continues, sharply reinforcing TSMC’s pricing power.
The results are clearly visible in the numbers. A Q1 gross margin of 66% is an exceptionally high level in the history of semiconductor foundries. NVIDIA’s TSMC purchase commitment is forecast to surge approximately sixfold from $16 billion to $95 billion within the next two years, meaning TSMC has secured negotiating power not merely as a contract manufacturer but as the critical bottleneck of the entire AI infrastructure. Some customers are prepaying years in advance to reserve capacity, reflecting the urgency around supply shortages.
The fact that SK Hynix has begun reviewing the introduction of Intel’s EMIB technology due to TSMC’s CoWoS supply shortage starkly illustrates how severe the TSMC supply bottleneck has become. At the same time, this represents a rare counterattack opportunity for Intel. However, Intel’s process yields and technical capabilities are still widely assessed as uncertain, and the market consensus is that TSMC’s monopolistic position is unlikely to be shaken in the short term. Ultimately, the advanced foundry market structure led by TSMC is likely to persist for at least several years, which is expected to translate into continued margin improvement and order expansion for TSMC.
3. Alibaba Cloud Forecasts MaaS Revenue 5-Year CAGR of 235% — AI-Related Revenue to Surge from 24B to 585.5B Yuan, Cloud Mix Shifting from 15% to 70%
• Core Source
“Citi forecast that Alibaba Cloud’s AI-related revenue will grow at an annual average of 90% over the next five years. AI-related revenue, currently at approximately 24 billion yuan, is expected to surge to 585.5 billion yuan by 2031, with its share of total cloud revenue expanding from 15% to 70%.”
“In particular, MaaS (Model-as-a-Service) revenue growth is forecast to reach an annual average of 235%.”
“MaaS-related revenue is forecast to grow at a CAGR of 235% to reach 438.6 billion yuan (USD 62.6 billion) by 2031, accounting for 53% of total cloud revenue.”
“Citi assessed that Alibaba is already positioned across four of the five layers of the AI industry value chain: energy, chips, infrastructure, models, and applications.”
• Expected Impact
The figures Citi presented in designating Alibaba Cloud as “China’s Google” suggest not merely a growth outlook but a fundamental transformation of the business structure. Non-AI cloud, which currently accounts for 85% of cloud revenue, is set to shrink to 30% by 2031, while the share of AI-related revenue flips from 15% to 70%. This means not simply adding AI features, but the business model itself being restructured around MaaS (Model-as-a-Service).
MaaS goes beyond renting cloud infrastructure to selling AI models themselves as a service. It commands higher unit prices, has the character of recurring revenue, and its customer lock-in effect is far stronger than infrastructure rental — making it a structural upgrade in terms of profitability. This is the backdrop to Citi’s presentation of the extreme figure of 235% CAGR.
Alibaba has already built a vertically integrated full-stack system encompassing its own AI chips (T-Head), cloud infrastructure (IaaS), PaaS, and MaaS, which is a structure that reduces dependence on external chips and clouds while maximizing margins. If the trend of Chinese companies relying on domestic AI stacks accelerates amid tightening U.S. export controls, Alibaba Cloud’s beneficiary position would grow even larger. Citi reconfirmed Alibaba as its top China AI pick, maintaining a price target of $205 on the ADR basis.
4. UBS Survey: Only 19% of Companies Have Deployed AI at Scale, Expectation-Reality Gap Continues to Widen — ROI Uncertainty and Demand Sustainability Arguments Collide
• Core Source
“According to a UBS survey, as of March this year, only 19% of companies had actually deployed AI at scale, and the pace of progress over the past two years has essentially been limited to linear growth.”
“A year ago, 84% of companies expected to complete large-scale AI adoption within the next 12 months, but only 5% actually achieved it.”
“This structural optimism bias has appeared repeatedly in every survey, and the gap between expectations and reality continues to widen. The main obstacles cited were unclear ROI, regulatory and compliance issues, system integration complexity, and talent shortages.”
“Semiconductor/memory momentum does not end when growth slows somewhat — it ends when the market withdraws its belief that supply shortages will continue.”
• Expected Impact
The UBS survey results expose a core contradiction embedded in the current AI investment cycle. A year ago, 84% of large companies expected to complete large-scale AI deployment within 12 months, but only 5% actually achieved it. The fact that the gap between expectations and reality recurs every year while actually widening is a structural problem.
The main obstacles are not technical. Unclear ROI, regulatory and compliance issues, system integration complexity, and talent shortages — organizational and institutional friction — are the core. This suggests that the transition from AI infrastructure demand at the hardware procurement stage to actual operational deployment is progressing far more slowly than expected.
However, this does not mean a slowdown in AI infrastructure demand itself. JP Morgan analyzes that only a visible slowdown in AI capex or credible evidence that memory supply is growing faster than expected would be the sole definitive ‘knockout blow’ to current semiconductor and memory momentum. In fact, big tech Capex guidance continues to be revised upward, showing that enterprise-level deployment delays are not preventing hyperscaler-level investment. Ultimately, the UBS survey results are best interpreted not as the disappearance of AI demand, but as a shift in the source of demand — that is, evidence of demand concentrating from corporate end-users toward hyperscalers. The simultaneous launch of dedicated AI deployment organizations by both OpenAI and Anthropic can also be seen as a move to bridge this gap.
5. OpenAI Sets $38B Cap on Revenue Sharing with Microsoft — Long-Term FCF Improvement and Expanded Flexibility for Cooperation with Google and Amazon
• Core Source
“Under the existing contract, OpenAI pays 20% of revenue to Microsoft. The total could reach up to $135 billion upon achieving long-term revenue targets through 2030.”
“Under the new contract, the 20% revenue sharing structure is maintained, but a cap is applied to total payments. The reported cap is $38 billion. OpenAI expense-accounts MS payments. Cumulative cost burden reduced.”
“Based on projected 2026 revenue of $30 billion, MS payments would be $6 billion (vs. the previous estimate of $4 billion based on deferral effects).”
“FCF deficit could also expand from the previous estimates of $25 billion in 2026 and $57 billion in 2027 to $27 billion in 2026 and $63 billion in 2027.”
“This move is expected to allow OpenAI to secure more flexibility to cooperate with other big tech companies such as Amazon and Google as it pursues an IPO in the future.”
• Expected Impact
The renegotiation of the revenue sharing agreement between OpenAI and Microsoft is on the surface a simple contractual adjustment, but within it lie two conflicting objectives simultaneously: short-term cash flow deterioration and long-term independence.
In the short term, the burden increases. Under the existing contract, some payments could be deferred until 2032, but under the new contract deferral is no longer possible, meaning 2026 MS payments rise from the previously expected $4 billion to $6 billion. The FCF deficit may expand to $27 billion in 2026 and $63 billion in 2027.
Over the long term, however, this is positive. By capping total distributions at $38 billion, a structure is created where even if OpenAI’s revenue grows beyond that level, the cost paid to MS no longer increases. Compared to having to pay up to $135 billion under the existing contract, the long-term FCF improvement effect is very significant. More importantly, this move reduces the constraints preventing OpenAI from cooperating with big tech players other than Microsoft such as Google and Amazon. This reads as a strategic move by OpenAI — with ARR currently estimated at approximately $35 billion — to maximize its valuation as an independent platform not beholden to any specific partner as it approaches an IPO.
6. Circle Unveils Agent Stack with USDC On-Chain Volume +263% — The Opening of the Stablecoin Platform Competition for AI Agent Payment Infrastructure Dominance
• Core Source
“USDC circulation reached $77 billion, up 28% year-over-year, and on-chain transaction volume surged approximately 263% year-over-year to about $21.5 trillion.”
“As Circle put forward Agent Stack, nanopayments, wallet, and developer tools, the story of ‘an economy where AI agents automatically pay using USDC’ has strengthened. Barron’s directly cited AI agent payment betting as the background for Circle’s stock surge.”
“Arc is a public blockchain specialized for institutional finance that uses USDC as its gas token. Circle’s goal is an ‘economic operating system’ encompassing payments, tokenized assets, and AI agent transactions while reducing dependence on external networks such as Ethereum and Solana.”
“ARC token presale: $222 million raised, network value of $3 billion on a fully diluted basis. Participation by major institutional investors including a16z crypto.”
“Number of CPN registered financial institutions: 136 (+36% QoQ).”
• Expected Impact
The core concept of the Agent Stack unveiled by Circle is an M2M (Machine-to-Machine) economy where AI agents directly pay for API fees, data costs, SaaS subscriptions, and more without human intervention. With a structure enabling micropayments as small as $0.000001 and gas-free transfers, the strategy is to be first to capture the real-time automatic payment market between AI agents that existing financial systems cannot cover.
The reason the market is paying attention is that USDC’s on-chain transaction volume has already surged 263% year-over-year to $21.5 trillion, numerically proving that the real-use base of stablecoins is materializing. Circulation of $77 billion (+28% YoY) and a 28% stablecoin market share demonstrate that USDC is functioning not as a speculative asset but as actual payment infrastructure.
The competitive landscape is also noteworthy. With Coinbase targeting the same market through its x402 payment standard based on HTTP 402 and the Base ecosystem, a platform standards war has begun over who will dominate the AI agent payment and market layer. The participation of a16z, BlackRock, Apollo, and ICE (parent company of the New York Stock Exchange) in the Arc token presale reflects the perspective of Wall Street institutions viewing this not as a simple crypto issue but as a next-generation financial infrastructure platform competition. If regulatory clarity progresses through CLARITY Act and similar legislation, the growth rate of the stablecoin-based AI agent payment market could accelerate further.
Comment [0]