Semiconductor Industry

Reports that Apple has shown interest in sourcing DRAM from China's CXMT would, if confirmed, represent one of the most significant supply-chain decisions in the semiconductor sector this decade. CXMT has been developing LPDDR5 chips competitive with Samsung and SK Hynix, and Apple qualifies its suppliers with extraordinary rigour, meaning even exploratory interest signals that CXMT's technology has crossed a threshold. The strategic consequence is severe for the US export control regime: Washington has spent three years restricting chip equipment sales to China specifically to prevent CXMT and peers from reaching this capability level. If Apple, a US-headquartered company, validates Chinese memory at scale, the political pressure on the Commerce Department to tighten its own allied companies' sourcing will intensify sharply. Samsung and SK Hynix, whose combined market share in LPDDR is well above 70 percent, should be concerned about the pricing leverage Apple would gain if CXMT becomes a credible third source.

Seoul has announced an 880 billion dollar investment programme in semiconductors and AI infrastructure, combining state incentives with commitments from Samsung and SK Hynix across advanced packaging, next-generation DRAM, and domestic AI compute capacity. To put the number in context: it is roughly three times the UK's entire annual government capital spending. The plan is a direct response to the US CHIPS Act and Taiwan's own expansion, and it accelerates the demand cycle for lithography and advanced packaging tools, which matters immediately for anyone holding ASML or ARM. The second-order risk is that Korean subsidies further compress margins for non-subsidised fabs trying to compete on cost, including the European foundry capacity the EU has spent four years trying to build.

Taiwanese authorities have raided Supermicro's Taiwan operations as part of a chip smuggling investigation, adding fresh legal exposure to a company that spent 2024 and 2025 clawing back credibility after an accounting scandal and a near-Nasdaq delisting. Supermicro is one of the primary assemblers of Nvidia GPU server racks, meaning any disruption to its Taiwan operations hits the AI infrastructure supply chain at a critical node. If the smuggling allegation involves restricted chips reaching sanctioned entities, the US Bureau of Industry and Security enters the picture and this escalates well beyond a Taiwanese regulatory matter. Watch whether Nvidia or its hyperscaler customers begin quietly diversifying server assembly contracts as a precaution.

A combined ten-year investment pledge of $1.3 trillion from Samsung and SK Hynix is the kind of number that makes the UK's semiconductor strategy look like a rounding error. The capital is targeted at advanced memory, logic, and AI chip production, with the implicit understanding that Korean fabs intend to hold their position in HBM supply regardless of what Sophon-style architectural disruption does to demand curves. For UK policymakers, the relevant benchmark is that the entire British industrial strategy budget is measured in single-digit billions. For institutional investors, the read is that Korean semiconductor equities are pricing in sustained pricing power in HBM that the Sophon PFG-1 story, if it scales, directly challenges. These two stories are running in parallel this week and they point in opposite directions for the memory supply chain.

The US government is committing $250 million to a chip manufacturing startup backed by a high-profile billionaire investor, continuing the CHIPS Act-era strategy of deploying public capital to accelerate domestic semiconductor capacity outside the established TSMC and Intel duopoly. The bet reflects a specific lesson from the Apple price story above: memory and logic chip shortages are now a consumer-facing problem, not an abstract supply-chain risk, and Washington has decided that tolerating that vulnerability is more expensive than subsidising the fix. The question every investor in European chip equipment companies should be asking is whether ASML and Infineon are supplying into the beneficiary's fab plan or being deliberately routed around.

Yesterday's global tech sell-off, which started in US sessions on 23 June, carried through Asian markets overnight with particular pressure on semiconductor and enterprise software names. The Oracle redundancy figure arriving alongside a risk-off session is a bad combination: it reinforces the narrative that AI investment creates concentrated winners and broad-based headcount cuts rather than the productivity-for-all story that justified elevated valuations. For UK investors with Asia-Pacific tech exposure, the relevant question is whether this is a rotation out of growth into defensives, or the beginning of a genuine earnings-expectations reset. One session does not answer that. But the Oracle data point makes it harder to dismiss.

Zhipu, the HSTECH index's standout performer this year, has rallied approximately 1,100% and attracted precisely one short call, which is either a sign of extraordinary conviction in Chinese AI's domestic substitution story or a market that has stopped asking hard questions about fundamentals. New US Commerce Department guidance is specifically targeting the overseas-subsidiary loophole that allowed Chinese-owned entities to access Nvidia Blackwell processors through offshore structures, and it applies to future acquisitions rather than existing deployments. The mechanism matters: Chinese AI firms with chips already installed keep running, but their ability to expand compute capacity at the frontier is being progressively squeezed with each enforcement round. The counter-thesis, visible in Zhipu's price, is that architectural workarounds and domestic chip progress will outpace the controls. That bet has looked reasonable for two years. Whether it survives the Blackwell-specific rules is the question that the single short seller is asking that no one else in the market currently is.

The SFC is distributing a HK$1 billion compensation pool funded by PwC Hong Kong for Evergrande minority shareholders, and it has refused to pause despite a judicial review filed by Evergrande's own liquidators, as coverage of the scheme confirms. The underlying misstatements were not marginal: per the SFC's own findings, Evergrande overstated FY2020 revenue by RMB350 billion, a 69% inflation, and reported a RMB31.4 billion profit in a year when the true result was a RMB19.9 billion loss. The liquidators' legal challenge is substantive, arguing the SFC had no freestanding statutory power to settle misconduct claims with an auditor, which is the AFRC's remit, and that the scheme may prejudice the estate's own potential claims against PwC. The precedent matters beyond the Evergrande carcass: this is Hong Kong's first instance of auditors of a defunct company being compelled to compensate shareholders via a regulator-brokered fund, and if it survives the judicial review it reshapes auditor liability calculus for every Big Four firm running China-exposed Hong Kong listings.

Hedgeye's Felix Wang has put a HK$407 fair value on Zhipu as a short, running directly against the momentum that has made it the hottest name on the HSTECH Index this year, and his core argument is that DeepSeek's V4 model has ignited a price war among Chinese AI firms that will destroy Zhipu's pricing power before its revenue base is big enough to absorb the compression, as Bloomberg's coverage of the move details. The counter-thesis, which is why the stock ran in the first place, is that US restrictions on Anthropic and other Western frontier models are redirecting Chinese enterprise demand toward domestic providers, making Zhipu a structural beneficiary of geopolitical decoupling rather than a cyclical AI play. Both can be true simultaneously: the demand tailwind is real, but a price war among five or six domestic Chinese AI firms competing for the same enterprise budget compresses the economics of that tailwind fast. Meanwhile, Sarvam AI has become India's latest AI unicorn at $234 million raised from HCLTech, Bessemer, and Khosla Ventures, and Jeff Bezos has backed UK-based CuspAI in a $400 million round focused on AI for physical-world simulation. The latter two deals point to where the next wave of differentiated AI capital is flowing: away from foundation model plays and toward domain-specific and scientific applications where defensibility is higher.

ETF flows are becoming a bigger driver of SK Hynix trading than the actual memory chip fundamentals. The Korean giant's share price has surged 230 percent year-to-date on AI memory demand, pushing it into the $1 trillion market cap club, but iShares MSCI South Korea ETF now holds almost 30 percent of its $23.9 billion assets in SK Hynix alone. When a single-stock leveraged ETF tracking SK Hynix surged 50 percent on a day the underlying stock fell 8 percent due to liquidity issues in its market-making system, it confirmed the tail is now wagging the dog. Fund flows into memory-themed products like Roundhill Memory ETF, which dedicates 27 percent of its $11.6 billion to SK Hynix, are creating feedback loops that amplify every move in the underlying stock. AI infrastructure demand is real, but when ETF mechanics matter more than earnings, someone is overpaying.

The country is capturing a rare convergence of AI infrastructure demand and geopolitical defense spending just as domestic policy support lags peers. Samsung pledged $228 billion for new semiconductor facilities while the government allocated $786 million for AI chip R&D over five years. South Korea's memory chip leaders dominate global DRAM with Samsung holding 40.7 percent market share and SK Hynix at 28.8 percent, positioning them as critical suppliers for AI data centers. Defense and shipbuilding orders are surging from rearmament and supply-chain realignment away from China. The challenge is policy coordination: analysts argue South Korea is under-supporting the sector versus US, Chinese, Japanese and Taiwanese programs, with semiconductor tax credits and AI legislation stalled. Corporate strength is outrunning state strategy at precisely the wrong moment.

The semiconductor giant's shares fell 15 percent in after-hours trading despite reporting 48 percent revenue growth and AI chip revenue that more than doubled to $10.8 billion. The sell-off came after management's forward guidance fell short of the most bullish investor expectations for AI growth acceleration. The broader market followed, with the Nasdaq falling 1.8 percent as investors questioned whether AI valuations had outpaced fundamentals. Broadcom's stumble illustrates how even strong AI results can trigger massive losses when growth trajectories disappoint hyped-up expectations.

C.C. Wei just pledged to raise employee profit-sharing by more than 30% after internal complaints about compensation failing to match record AI-driven earnings. TSMC's 2025 revenue hit $122.4 billion, up 36%, with AI chips now commanding 58% of sales versus 51% last year, according to earnings reports. The move signals how tight the semiconductor talent market has become as TSMC races to build six Arizona fabs while ramping 3nm production. When your margins are 67%, keeping engineers happy becomes a strategic imperative.

ASML jumped 8% to a record $500 billion market value as AI optimism spreads beyond foundries to toolmakers and the broader semiconductor supply chain. While TSMC's guidance for 52-56 billion capex and 30% revenue growth validates continued AI demand, emerging market fund managers are diversifying into cheaper Asian AI plays. Valuations show the opportunity: Nvidia trades at 35x forward earnings while TSMC sits at 18x and Alibaba at 13x. The rotation reflects investor confidence that AI infrastructure spending above $1 trillion will support a multilayer ecosystem, not just the obvious winners.

Nvidia posted record quarterly revenue of $81.6 billion, up 85% year-on-year, as data center sales hit $75.2 billion on explosive demand for AI agents and generative AI workloads. The company has cycled past last year's $4.5 billion inventory charge on China-targeted H20 chips, with gross margins recovering to the mid-70% range as Blackwell platform sales scale. Revenue has grown 10x in three years, reshaping semiconductor power dynamics and making Nvidia the de facto standard for AI infrastructure from hyperscalers to enterprises.

Nvidia raised its quarterly dividend to $0.01 per share, maintaining an annual yield around 0.02% that keeps the stock firmly in growth territory despite record earnings. The company returned $41.1 billion to shareholders in fiscal 2026, overwhelmingly through buybacks rather than dividends, with $58.5 billion remaining under its repurchase authorization. Even symbolic dividend increases are being read as signals that management sees fewer reinvestment opportunities at previous returns, explaining why shares pulled back despite beats across revenue, margins, and guidance.

South Korea's Kospi surged over 150% in the past year on AI semiconductor demand, but the won hit 1,449 per dollar as Korean investors bought more US AI stocks than foreigners bought Korean chips. Semiconductor exports jumped 173% year-on-year to $31.9 billion, generating a $23.8 billion trade surplus, yet the currency remains one of Asia's weakest. KB Kookmin Bank's Moon Jung-hee attributes this to AI investment flows: domestic demand for US mega-cap AI plays exceeded foreign buying of Korean memory names, creating net dollar demand despite export windfalls.

Analog Devices is in advanced talks to acquire AI power management startup Empower Semiconductor for about $1.5 billion, marking a massive premium for specialized data center voltage regulation technology. Empower makes integrated voltage regulators that sit closer to AI accelerators and GPUs, enabling faster power management as workloads change dynamically. The deal would be ADI's largest acquisition since the $21 billion Maxim Integrated purchase, reflecting how AI infrastructure spending has elevated power delivery from a commodity to a strategic differentiator. For a private startup with undisclosed revenues, $1.5 billion suggests ADI is paying primarily for customer relationships with hyperscalers and the intellectual property needed to solve AI's power density challenges.

SkyeChip's IPO demand signals Asia wants its own NVIDIA proxies. The Penang-based chip designer drew 95x oversubscription for its public tranche, the largest retail demand since Petronas Chemicals in 2010. Malaysia's 2026 IPO proceeds are tracking toward $1.8 billion, the highest in 13 years, driven by tech listings and REIT spin-offs. Khazanah, EPF and other sovereign funds anchored the deal, betting on Malaysia's push to move beyond back-end semiconductor assembly into higher-value AI chip design. The frenzy mirrors Samsung's $1 trillion valuation spike as regional markets chase AI infrastructure plays.

The three companies controlling 90% of global memory supply just discovered what monopoly pricing looks like. Micron's shares jumped 38% last week after executives said demand is "nowhere close" to matching supply, while Samsung reported a 90% price increase in Q1 alone. DRAM contract prices are projected to surge 63% this quarter, the steepest jump in a decade. The math is brutal: AI servers need 8-10 times more memory than traditional servers, but new fabs won't come online until 2028. Apple is already warning about margin pressure, Sony raised PlayStation prices by $150, and smartphone shipments are forecast to drop 31% as manufacturers ration chips for higher-margin AI applications.