TSMC (TSM) on Solana

SoftBank Group disclosed in its Q1 2026 results that it sold 71.5% of its Taiwan Semiconductor Manufacturing (TSMC) holding while simultaneously initiating a new stake in Capital One. The move marks a sharp reallocation away from one of the world's largest chipmakers and toward U.S. financials, with SoftBank reporting Q1 revenue of ¥2,019.59 billion alongside net income of ¥347.33 billion, which eased compared to the prior period.

The TSMC exit is drawing investor attention to how SoftBank's asset sales and portfolio reshaping feed through to its earnings, cash generation, and capacity to support buybacks and dividends. Analysts are divided on whether the shift reinforces concerns about SoftBank's ability to fund ongoing AI investments, or whether it signals a deliberate pivot in strategic focus — a question that may prompt near-term revisions to institutional outlooks on the Japanese conglomerate.

TSMC reported July 2026 revenue of 467.58 billion New Taiwan dollars (~$14.5 billion), a 44.7% jump from the same month a year earlier, extending the record-breaking pace set in its most recent quarterly earnings. The figure comfortably exceeded the company's own full-year guidance of over 40% revenue growth and follows a Q2 2026 result of $40.2 billion — up 33.7% year-over-year — in which net income surged 77.4%. High-performance computing, the segment dominated by AI accelerator silicon, represented 66% of Q2 wafer revenue, with advanced-node technologies accounting for 77% of the total.

CEO C.C. Wei described AI demand as "extremely robust," and TSMC has raised its 2026 capital expenditure plan to $60–$64 billion to keep pace. The company's Q3 2026 revenue guidance of $44.6–$45.8 billion implies continued sequential growth. TSMC manufactures the most advanced chips for Nvidia and Google, making its monthly revenue disclosures a real-time gauge of the AI infrastructure build-out. The sustained acceleration reinforces the view that semiconductor supply chains remain under significant strain from hyperscaler and AI lab spending, with no sign of demand cooling in the near term.

TSMC and Sony plan to invest approximately $6.3 billion in a joint venture to produce high-performance image sensors at Sony Semiconductor Solutions' existing facility in Kumamoto, Japan. Sony will hold a 60% stake in the venture with TSMC taking 40%, and the two companies expect to finalize the investment agreement in the coming months. Mass production is targeted to begin as early as 2029.

The partnership pairs Sony's position as the world's largest image sensor maker with TSMC's advanced semiconductor manufacturing capabilities. Image sensors are a key component in smartphones and automotive systems, and the collaboration is aimed at developing next-generation imaging technology ahead of rising demand from both markets.

TSMC reported Q2 2026 revenue of $40.2 billion, up 36% year-over-year, with a 67.7% gross margin, while Intel posted $16.13 billion in the same period. Intel's foundry division lost $2.1 billion despite 31% revenue growth, underscoring the structural gap between the two companies' manufacturing operations. TSMC is funding a $52–56 billion 2026 capital expenditure program from operating cash flow; Intel, by contrast, is relying on a $5 billion equity investment from NVIDIA and $8.9 billion in CHIPS Act support to finance its expansion.

On the technology front, TSMC commercially shipped 2nm chips for the first time in Q2, with advanced nodes at 7nm and below now representing 77% of total wafer revenue. Intel's realistic near-term goal, analysts suggest, is proving yield maturity on its 18A process and winning secondary tile orders rather than displacing TSMC in AI silicon. Intel's Data Center and AI segment grew 59% to $6.26 billion, partly through Xeon 6+ adoption as the host CPU in NVIDIA's DGX Rubin systems — a design win that nonetheless keeps Intel positioned as a component supplier within TSMC-dependent AI infrastructure.

J.P. Morgan said Wednesday that the recent 25–30% drawdown in Asian technology stocks — the third major correction since the AI-driven upcycle began in late 2022 — has not altered its fundamental outlook for the sector, and that TSMC specifically remains a central beneficiary of the next phase of AI infrastructure buildout. The bank stated it sees "no fundamental indicators that signal meaningful weakness in the next 6–12 months," pointing to frontier AI models continuing to improve on a multi-month cadence and sustained demand for AI inference across both proprietary and open-source deployments.

Within that thesis, J.P. Morgan highlighted advanced packaging as an area of sharp near-term growth, noting that 2.5D packaging is going mainstream while the 3D packaging investment cycle is just beginning at TSMC. That assessment frames the current share-price weakness as a valuation reset rather than a signal of deteriorating demand for TSMC's most capacity-constrained and margin-accretive product lines.

TSMC is developing an internal chip-packaging technology described as "EMIB-like," borrowing directly from Intel's Embedded Multi-die Interconnect Bridge approach — a significant strategic pivot for a company that built its dominance on pure-play foundry manufacturing. The move is driven by a concrete bottleneck: TSMC's CoWoS-L advanced packaging capacity is sold out through 2026 and into 2027, with lead times stretching to 78 weeks, opening a window for Intel to court major customers. Nvidia is already evaluating Intel's EMIB for a future processor combining four GPUs into a single package, and Intel has secured packaging orders from Nvidia, Google, and OpenAI.

The strategic concern for TSMC runs deeper than packaging revenue. As one framing in the coverage puts it, "packaging orders are rarely just packaging orders" — a customer who trusts a foundry to assemble its chip is a customer that can also be pitched on manufacturing it. By developing its own EMIB-like capability, TSMC is defending the customer relationships that anchor its foundry leadership, not just chasing a new revenue line. The episode signals a broader shift in the semiconductor competitive landscape: with transistor scaling delivering diminishing differentiation, advanced packaging is emerging as the next arena where a manufacturing edge materially matters.

Cathie Wood's ARK Invest acquired roughly $20 million in TSMC (TSM) shares on July 29, 2026, spreading the purchase across four ETFs — ARKK, ARKQ, ARKW, and ARKX — after the stock shed approximately 12% over six consecutive sessions. The buy came on the heels of TSMC's latest quarterly earnings, which beat market expectations and prompted management to raise its full-year revenue growth guidance to roughly 40%, up from an earlier 30% estimate.

TSM edged 0.3% higher in premarket trading on July 30, snapping the losing streak. The stock remains up about 24% year-to-date, supported by demand from major AI infrastructure spenders including Alphabet, Amazon, and Microsoft, which have collectively committed $725 billion in AI-related capital expenditure for 2026. Thirty-four of 35 analysts tracked by Koyfin rate TSM a buy, with the lone dissenter at hold.

TSMC shares are trading roughly 27% below the consensus analyst target of $534.51, with the stock down about 9.3% over the past 30 days as two distinct concerns weigh on sentiment. The first is the emergence of Chinese-developed ultraviolet etching technology that investors worry could gradually erode TSMC's leadership in advanced process manufacturing. The second is broader uncertainty around how long the current AI-driven spending cycle can sustain demand for high-end chips — a question with direct implications for TSMC's capacity expansion plans.

Analysts note that a high proportion of non-cash earnings complicates profit-quality assessments during periods of sector stress, adding another layer of caution. The stock is currently trading near estimated fair value by some measures, meaning any re-rating upward depends on TSMC demonstrating that Chinese toolmakers remain well behind in practical capability, and that AI accelerator orders remain durable beyond the current buildout wave.

Political pressure from the Trump administration to produce AI chips domestically is creating a tangible drag on TSMC's profitability. The company has committed over $200 billion to US semiconductor manufacturing since Trump returned to office — including a recently announced additional $100 billion for Arizona fabs — but operating in the US costs 20–50% more than producing the same chips in Taiwan. TSMC has warned investors of a 2–4% gross margin dilution over several years as its American facilities ramp up, with higher production costs expected to be partially passed on to customers.

The financial pressure arrives even as TSMC's underlying business is booming: Q2 2026 profit surged 77% on AI-driven demand and the company raised its full-year revenue growth forecast to over 40%. TSMC plans up to 12 fabs and an R&D center in Arizona, cementing its role as the primary supplier of advanced AI silicon in the US market. Investors are weighing whether the long-term strategic value of domestic manufacturing — reduced geopolitical risk and access to US government incentives — offsets the near-term margin compression that comes with it.