News related to
chip

The Tokyo-based maker of NAND flash memory and solid-state drives has become one of the most closely watched stocks on the Tokyo Stock Exchange after a sharp rally tied to demand for data-centre storage, AI servers and high-performance memory systems. Its market capitalisation has climbed into the tens of trillions of yen, placing it near Toyota and behind SoftBank Group in a reshuffled ranking that would have seemed unlikely when Kioxia priced its initial public offering at 1,455 yen a share in December 2024.

The shift marks a wider challenge to the long-standing dominance of Japan’s manufacturing champions. Toyota, the global carmaker that symbolised the strength of Japan Inc. for decades, has been displaced from the top position by SoftBank during the latest AI-led surge, while Kioxia has risen rapidly from a former Toshiba memory unit backed by Bain Capital into a central player in the market’s technology rotation.

Kioxia’s climb reflects investor expectations that memory will remain a bottleneck in AI infrastructure. Generative AI systems require vast amounts of storage and high-speed data movement, pushing demand beyond the older cycle driven mainly by smartphones and personal computers. Data-centre operators, cloud companies and chip designers are competing for components needed to train and run large AI models, giving memory producers stronger pricing power after years of volatility.

The company’s earnings have strengthened sharply. Kioxia has projected operating profit of about 1.3 trillion yen for the April-June quarter, a figure that puts it among the most profitable listed companies in Japan. Its latest annual results showed strong momentum in revenue, margins and cash generation, helping ease earlier concerns about debt and cyclicality. Rating upgrades to investment grade have added to the perception that the balance sheet is improving at a time when investors are rewarding semiconductor exposure.

Kioxia’s origins give its rise broader industrial significance. The company traces its roots to Toshiba’s memory business, which was sold to a Bain-led consortium in 2018 after Toshiba’s financial crisis. The group was renamed Kioxia in 2019, drawing on the Japanese word for memory and the Greek word for value. Its public listing in 2024 valued it below 800 billion yen, after earlier IPO attempts were delayed by weak valuation conditions and pressure in the memory market.

That cautious debut contrasts sharply with its current status. Shares have multiplied as investors reassessed the role of NAND flash in AI systems, especially as large-scale data centres require higher storage density, faster access and more energy-efficient hardware. Kioxia has also been developing advanced BiCS FLASH technology and promoting next-generation SSDs for AI storage infrastructure.

The rally has also lifted questions over valuation risk. Memory chips remain one of the semiconductor industry’s most cyclical segments, with prices exposed to shifts in capital spending, inventory levels and customer demand. A sharp rise in market value leaves Kioxia vulnerable if hyperscaler investment slows, if competitors add capacity faster than expected, or if customers push back against higher pricing.

Competition is intense. Samsung Electronics, SK Hynix, Micron Technology and SanDisk remain key players across memory markets, with SK Hynix holding a dominant position in high-bandwidth memory used in AI accelerators. Kioxia’s strength is more closely tied to NAND flash and SSDs, where demand from AI infrastructure has broadened the market but does not eliminate the risk of oversupply once new capacity comes on stream.

Toyota’s relative decline in the rankings does not point to operational weakness alone. The carmaker remains one of the world’s largest manufacturers by sales and a major profit generator, but investors have been rotating away from traditional industrial leaders towards companies with direct AI exposure. Concerns over electric-vehicle competition, currency movements and the cost of future technologies have weighed on auto valuations, even as Toyota continues to command global scale in hybrid vehicles and supply-chain management.

SoftBank’s ascent has further highlighted the changing market narrative. Its exposure to Arm and AI-related investments has drawn investors seeking Japan-based access to global AI infrastructure. Together, SoftBank and Kioxia now represent a market story centred on chips, data centres, software platforms and computing capacity rather than the export-led industrial model that long defined Japan’s equity market.

Foreign investor flows have amplified the shift. Japan’s equity market has benefited from corporate-governance reforms, stronger capital returns and renewed interest in companies linked to the AI supply chain. The Nikkei 225 has reached record territory, though the rally has been uneven, with gains concentrated in a limited group of technology-linked names while broader market performance has been more mixed.

Kioxia’s next test will be whether it can convert favourable pricing and sold-out capacity into durable earnings. Investors will look for clearer shareholder-return plans, disciplined capital expenditure and evidence that AI storage demand can remain strong beyond the current investment wave. The company’s contemplated US share listing would widen access to global technology investors and could further raise its profile if market conditions remain supportive.

The shift reflects a sharp change in investor preference after months of underperformance by Hong Kong equities against mainland peers. Mainland investors sold about HK$3.6 billion of Hong Kong shares through the trading links last month, the first monthly net outflow since June 2023 and one of only a small number of monthly selling episodes since the southbound channel became a key route for mainland money into the city’s market.

The move is striking because southbound buying has been one of the strongest supports for Hong Kong stocks over the past year. Mainland funds poured into high-dividend banks, telecoms groups, energy companies and internet platforms during earlier phases of the rally, helping offset caution among global investors. That support is now weakening as domestic funds chase companies tied to AI infrastructure, chips, optical modules, software and computing power.

Onshore AI shares have delivered a stronger performance than Hong Kong’s technology benchmarks this year. The CSI Artificial Intelligence Index has risen sharply, while the Hang Seng Tech Index has lagged, weighed down by concerns over platform margins, heavy spending on AI models and pressure from subsidy battles in e-commerce and online services. The divergence has encouraged active money managers and retail traders to reassess whether Hong Kong-listed internet giants still offer the most attractive exposure to China’s technology story.

The change in flows also comes as Beijing’s industrial policy continues to favour advanced manufacturing, domestic semiconductors, large language models and high-end computing. Companies linked to AI servers, data centres, optical communications, robotics and chip design have become favoured market proxies for China’s effort to reduce dependence on foreign technology. That policy backdrop has helped support valuations even as questions persist over earnings visibility, capital expenditure demands and competition.

Cambricon Technologies, Zhongji Innolight, Kingsoft Office, iFlytek and other mainland-listed technology names have been among the companies watched closely by investors seeking domestic AI exposure. Hong Kong still hosts major technology groups such as Tencent, Alibaba, Baidu, Meituan and Xiaomi, but the market’s heavyweight structure makes it more exposed to mature platform businesses than to the faster-growing AI hardware and infrastructure supply chain favoured by onshore traders.

Hong Kong’s role as a fundraising centre remains intact despite the outflow. The city has continued to attract listings from technology, healthcare, robotics and semiconductor-related companies, while mainland firms still view it as the most practical offshore venue for raising international capital. Several AI and chip-linked companies are preparing or considering Hong Kong share sales, suggesting the city remains central to China’s capital market architecture even as secondary-market money rotates elsewhere.

The weakness in southbound appetite also reflects broader caution over Hong Kong valuations after a strong earlier advance. Investors who bought the market for dividends and policy support are locking in gains while seeking higher-beta opportunities at home. The shift does not necessarily point to a structural rejection of Hong Kong stocks, but it does show that mainland money is becoming more selective after helping drive earlier momentum.

Regulatory pressure on offshore investment channels may also be influencing behaviour. Authorities have tightened scrutiny of unauthorised cross-border trading accounts and offshore brokerage activity, reinforcing the appeal of regulated domestic routes. While Stock Connect remains an approved mechanism, wider efforts to keep capital within supervised channels have added another layer of sensitivity around outbound allocations.

Market analysts say the key question is whether the AI-led onshore rally can broaden beyond a narrow group of highly valued companies. China’s AI supply chain is benefiting from state support, falling model costs and growing enterprise adoption, but investors face risks from crowded positioning, uneven profitability and rapid technological shifts. Several AI-linked companies still trade more on expectations than on proven earnings, leaving the sector vulnerable to volatility if sentiment turns.

The Provisions on the Protection of Trade Secrets, issued by the State Administration for Market Regulation under Order No. 126, came into force on 1 June 2026 and replace rules first introduced in 1995. The overhaul gives regulators clearer authority to handle disputes involving digital assets that now sit at the centre of corporate value, from AI model architecture and training datasets to platform recommendation systems, software code and commercially valuable operational data.

The change marks a significant shift in China’s treatment of intangible technology assets. Earlier rules were drafted before cloud computing, AI systems and large-scale data operations became central to business competition. The new framework defines trade secrets as technical or business information that is not publicly known, has commercial value and has been subject to confidentiality measures by the rights holder. Technical information now expressly covers data, algorithms, computer programs and code, while business information can include management, sales, finance, planning, customer and operational data.

The timing gives the regulation wider strategic weight. China and the United States are locked in a prolonged technology rivalry involving semiconductor controls, restrictions on advanced computing hardware, limits on investment in sensitive sectors and scrutiny of cross-border technology transfers. Washington has tightened export controls on high-end chips and chipmaking equipment, while Beijing has expanded its own legal toolkit to protect critical technology, secure data flows and prevent overseas transfer of capabilities it regards as strategically important.

The new rules do not create a separate trade secret law, but provide more detailed administrative guidance under China’s Anti-Unfair Competition Law. They clarify how regulators should assess whether information is secret, whether it has commercial value, what counts as reasonable confidentiality measures and how infringement should be investigated. For companies, the central message is that protection will depend not only on the value of the information, but also on whether internal controls, access limits, confidentiality agreements, digital safeguards and evidence trails are in place.

The provisions also recognise that unsuccessful research can have commercial value. Stage-by-stage technical results, failed experimental data and development paths that save time or cost in future research may qualify for protection if they meet the legal conditions. That is especially relevant for AI labs, pharmaceutical developers, electric vehicle makers, chip design firms and advanced manufacturing groups, where failed trials can reveal important technical boundaries and prevent competitors from repeating expensive mistakes.

Regulators have been given practical enforcement tools, including powers to inspect business premises, question individuals, copy documents, examine electronic records and take action against materials linked to suspected infringement. Penalties can include orders to stop unlawful conduct, confiscation of illegal gains and fines that may reach CNY5 million in serious cases. The framework also covers improper acquisition, disclosure, use or permission to use trade secrets obtained through theft, bribery, fraud, coercion, electronic intrusion or breach of confidentiality obligations.

For multinational companies operating in China, the new rules create a mixed picture. Stronger protection could help firms defend proprietary data, software systems and industrial processes in a market where trade secret disputes have often been difficult to pursue. At the same time, wider protection for data and algorithms may increase compliance complexity, particularly in joint ventures, outsourced research, cloud services, employee mobility, cross-border collaboration and licensing arrangements.

Technology companies are likely to face tougher expectations over documentation. Firms seeking protection will need to show that the information was not easily accessible, carried economic value and was actively protected. Generic claims that an algorithm or dataset is confidential may not be enough. Companies will be expected to classify sensitive assets, restrict access by role, track downloads and transfers, maintain logs, use encryption or watermarking where appropriate, and ensure that employment and partnership contracts define confidentiality obligations with precision.

The rules also sit alongside China’s broader data and security regime, including laws on data security, personal information protection, cybersecurity and export control. That overlap could make enforcement more powerful but also more complicated. A dataset may be a trade secret, a regulated data asset and a cross-border transfer concern at the same time. Businesses handling AI training material, user behaviour data or industrial performance data will need to assess those obligations together rather than treating trade secret protection as a standalone issue.

Shares in Masayoshi Son’s technology investment group surged in Tokyo trading, lifting its market value to about ¥47.2 trillion, or roughly $296 billion, ahead of Toyota’s ¥45.7 trillion after the carmaker’s stock fell nearly 5 per cent. The move placed SoftBank at the top of Japan’s corporate valuation table for the first time in more than two decades and added momentum to a Nikkei 225 rally that pushed the index above 67,000.

The change is more than a stock-market milestone. It reflects how the AI boom has reshaped capital allocation in Japan, where investors are assigning higher value to companies with exposure to semiconductors, data centres, power infrastructure and generative AI platforms. Toyota remains one of the world’s largest and most profitable carmakers, but its valuation is now being weighed against a slower earnings outlook, tariff pressure, currency swings and the heavy cost of electrification.

SoftBank’s advance has been driven by a powerful combination of portfolio gains and expectations that its AI-linked holdings could generate fresh liquidity. The group’s exposure to Arm, the British chip-design company it took public in 2023, has become a key pillar of its market value. Arm’s technology is used widely across smartphones, cloud computing and increasingly AI-related chips, giving SoftBank a central position in the semiconductor supply chain.

OpenAI has become the second major driver of investor attention. SoftBank has committed tens of billions of dollars to the ChatGPT developer and is positioning itself as a core financier of AI infrastructure. Expectations that OpenAI could move towards a public listing have intensified interest in SoftBank’s asset base, even as analysts continue to debate how much of the group’s valuation depends on private-market pricing and future monetisation.

The latest surge followed SoftBank’s plan to invest up to €75 billion in AI infrastructure in France, including data centres and power-intensive computing capacity. The first phase is expected to involve about €45 billion of investment to build 3.1 gigawatts of capacity by 2031, with further expansion potentially taking the total to 5 gigawatts. The project places France at the centre of Europe’s push to develop sovereign AI infrastructure, helped by its nuclear-heavy power system and available industrial sites.

SoftBank is also moving to create public-market vehicles around AI infrastructure. SB Energy, its power and data-centre platform, is preparing for a potential US listing and has been linked to a valuation above $50 billion. The company is involved in large-scale power projects designed to support AI workloads, including a 1.2-gigawatt data-centre development in Texas. A separate robotics venture, Roze, is also being prepared for a possible listing, highlighting SoftBank’s attempt to build an ecosystem around energy, construction automation and computing capacity.

Toyota’s fall in ranking does not signal operational weakness. The carmaker reported operating income of ¥3.8 trillion for the financial year ended March 2026, supported by vehicle sales and pricing power despite tariff costs and rising expenses. It remains a global force in hybrid technology, production efficiency and supply-chain management. However, investors have become more cautious about its growth outlook as the auto sector faces higher regulatory costs, slower battery-electric vehicle adoption in some markets, and intensifying competition from China-based manufacturers.

The contrast between SoftBank and Toyota captures a broader reordering of Japan’s equity market. AI-linked names, chip suppliers and infrastructure companies are drawing strong inflows, while parts of the auto sector and older industrial groups face questions over margin durability. Kioxia, the memory-chip maker, has also climbed sharply this year, showing how investor enthusiasm has spread beyond SoftBank into companies tied to storage and high-performance computing.

The rally has strengthened the profile of Son, whose investment record has swung between spectacular gains and painful losses. SoftBank’s Vision Fund suffered heavy write-downs after the collapse in technology valuations in 2021 and 2022, while earlier bets such as WeWork damaged confidence in the group’s risk controls. Its comeback has been powered by a more concentrated AI narrative, but that also leaves the shares vulnerable if expectations around OpenAI, Arm or infrastructure listings are revised lower.

The memorandum of understanding, signed by the Government of Odisha, Intel and 3DGS, covers an advanced packaging glass-core substrate manufacturing facility in the Bhubaneswar-Khurda region. The project is expected to focus on substrates used in next-generation chip packaging, a part of the semiconductor supply chain that has gained strategic importance as artificial intelligence, high-performance computing and advanced data-centre processors demand larger, faster and more power-efficient packages.

The agreement places Odisha in a stronger position within India’s semiconductor map, which has so far been led by assembly, testing, marking and packaging projects, outsourced semiconductor assembly facilities, compound semiconductor proposals and early-stage fabrication investments. The planned plant would add a specialised layer to that ecosystem by targeting glass-core substrates and high-density interconnect technologies, areas that are increasingly viewed as critical for advanced packaging.

Intel’s role is expected to centre on technology know-how, process expertise and ecosystem support, while 3DGS brings specialisation in glass-based integration platforms. 3DGS, a US-based company, has developed glass-centric technologies for radio-frequency, photonics and semiconductor packaging applications. Its involvement gives the Odisha plan a sharper focus than conventional electronics manufacturing projects, as glass-core substrates remain an emerging technology rather than a mature commodity segment.

Glass substrates are attracting industry attention because conventional organic substrates face physical limits as chip packages become larger and denser. Advanced processors now often combine multiple chiplets, memory stacks and interconnect layers in one package. That increases pressure on the base material to carry more power and data connections while reducing warpage, signal loss and thermal stress. Glass offers better dimensional stability, finer interconnect potential and the ability to support larger package sizes, though yield, tooling and commercial-scale production remain challenges.

Intel has publicly positioned glass substrates as a key packaging technology for the second half of this decade. The company has argued that glass could help extend system-level performance gains beyond the limits of transistor scaling, particularly for artificial intelligence accelerators and high-performance computing chips. For India, participation in this segment could move the country beyond back-end assembly into a more advanced part of the global chip supply chain.

The Odisha project is also politically significant. The signing was witnessed by Union electronics and information technology minister Ashwini Vaishnaw and Odisha chief minister Mohan Charan Majhi, alongside senior executives from Intel and 3DGS. The state has been pushing to attract electronics, data centre and semiconductor-linked investments as part of a broader industrial strategy aimed at diversifying beyond metals, mining and ports.

The proposed facility is expected to be developed over several years, with industry estimates placing the build-out period at five to six years. Employment projections point to high-skilled roles across engineering, materials science, process control, quality systems and equipment maintenance, while indirect jobs could emerge through chemicals, gases, logistics, clean-room services and precision component suppliers.

The investment comes as India seeks to reduce dependence on overseas semiconductor supply chains after pandemic-era shortages exposed vulnerabilities in electronics, automotive and defence manufacturing. The Union government’s semiconductor programme has already backed projects by Tata Electronics, Micron Technology, CG Power and Kaynes, while several states are competing to attract design, packaging and manufacturing facilities. Gujarat and Assam have emerged as early beneficiaries, while Odisha’s Intel-3DGS agreement signals a broader geographic spread of the chip ecosystem.

For Intel, the deal fits into a wider effort to strengthen its foundry and advanced packaging credentials at a time when the global chip industry is being reshaped by artificial intelligence demand. The company has been investing in packaging platforms such as EMIB and Foveros, while also promoting glass substrates as a future enabling technology. Its broader turnaround remains under scrutiny because of competitive pressure from Nvidia, AMD, TSMC and Samsung, but advanced packaging offers a route to relevance in custom accelerators and multi-die systems.

The project will still have to pass execution tests before it becomes a production anchor. Semiconductor substrate manufacturing requires tight process control, specialised equipment, stable utilities, skilled technical labour and reliable supply of materials. Glass-core technology also needs commercial validation at scale, as the industry has not yet shifted from established organic substrate supply chains in volume applications.

The California-based server maker said the equipment had first been sold through an authorised reseller before being deceptively acquired and targeted for illicit diversion. Its statement framed the case as a successful intervention rather than a failure of internal controls, stressing that it had worked with Taiwan authorities to stop the movement of high-value systems used in artificial intelligence data centres.

The case has drawn attention because Supermicro sits at the centre of the AI infrastructure boom, supplying servers that integrate advanced graphics processors and networking components used to train and run large AI models. Such systems have become strategically sensitive as Washington, Taipei and allied governments tighten oversight of technology flows to mainland China, Hong Kong and Macau.

Taiwan prosecutors have been investigating suspects accused of falsifying export documents for high-end AI servers containing Nvidia chips. The servers were allegedly routed through Japan as part of an attempt to move restricted technology onwards to China-linked buyers. The seizure of 50 systems, though small compared with global AI server demand, is significant because it shows authorities are now pursuing hardware diversions beyond the better-known routes through Southeast Asia.

Supermicro’s response is also shaped by a separate and larger US criminal case that has placed the company under scrutiny. In March, US prosecutors charged three people linked to Super Micro Computer, including co-founder Yih-Shyan “Wally” Liaw, in an alleged conspiracy to divert about $2.5bn worth of AI server technology to China in violation of export controls. Liaw and another defendant were arrested, while a third defendant was listed as a fugitive. Super Micro Computer itself was not named as a defendant.

That case alleged a sophisticated scheme involving shell companies, false shipping documents, repackaging and dummy servers designed to mislead compliance checks. Prosecutors said controlled AI server technology was moved from the United States to Taiwan before being redirected through intermediary locations to reach Chinese customers. The allegations intensified questions over how manufacturers, distributors and resellers monitor the final destination of advanced hardware once it leaves controlled channels.

Supermicro has denied institutional wrongdoing and has said the conduct alleged in the US case violated company policies. The latest Taiwan-linked statement attempts to reinforce that message by presenting the firm as a partner to investigators. It thanked law-enforcement and legal officials, while saying industry and government must work together to strengthen safeguards and improve supply-chain visibility.

The timing matters for investors. Supermicro has been one of the most closely watched hardware suppliers during the AI build-out, benefiting from surging demand for rack-scale servers used by cloud providers, enterprises and sovereign AI projects. That growth has made export-control compliance a material business risk. A single enforcement case can raise concerns about sales screening, reseller audits, customer identification and shipment tracking, even when the company is not accused of direct wrongdoing.

Nvidia’s position adds another layer to the issue. Its most advanced AI processors remain central to global data-centre expansion, and restrictions on sales to China have forced chipmakers and server companies to navigate a shifting regulatory environment. Nvidia has repeatedly said it follows export rules and does not support unlawfully exported products. Its chief executive, Jensen Huang, has also underlined that server vendors must maintain their own compliance standards when handling systems containing restricted chips.

The Taiwan action highlights a broader enforcement shift. Authorities are no longer focused only on chip shipments; fully assembled AI servers are now a priority because they can be more difficult to trace after integration. Servers may contain restricted processors, specialised networking, memory and cooling systems that together create the computing power needed for frontier AI workloads. Once such systems are broken into multiple shipments or passed through resellers, detecting diversion becomes more complex.