Internet of Things - Company and Research Focus

The installation of new CEO Lip-Bu Tan at Intel was supposed to be a turning point. But as it turns out, it is leaving behind a new wake of uncertainty that is prompting concern. Apparently, Tan has a wide range of investments, many of which are in China…and some of which are even connected to the Chinese military. This was bad news as far as investors were concerned, and Intel shares plummeted over 7.5% in Thursday afternoon’s trading. Given that Tan put up $25 million of his own money to invest in Intel when he took the CEO slot, it should come as little surprise that he had substantial investments to begin with. Tan’s investments, reports noted, included “hundreds of Chinese tech firms.” The biggest surprise, though, was that at least 8 of these firms had direct ties to the People’s Liberation Army, the Chinese military. A review from Reuters found that Tan actually controls +40 Chinese companies outright. He has minority stakes in +600 beyond that as well. And this, investors believe, may “complicate” the notion of getting Intel back up and running. Andrew King with Bastille Ventures noted “The simple fact is that Mr. Tan is unqualified to serve as the head of any company competing against China, let alone one with actual intelligence and national security ramifications like Intel and its tremendous legacy connections to all areas of America’s intelligence and the defense ecosystem.”

The penalties announced by Trump on Wednesday triggered a plunge in world financial markets and drew condemnation from other leaders reckoning with the end of a decades-long era of trade liberalization. In Japan, one of United States' top trading partners, Prime Minister Shigeru Ishiba said that the tariffs had created a "national crisis" as a plunge in banking shares on Friday set Tokyo's stock market on course for its worst week in years. Investment bank JP Morgan said it now sees a 60% chance of the global economy entering recession by year end, up from 40% previously. The U.S. tariffs would amount to the highest trade barriers in more than a century: a 10% baseline tariff on all imports and higher targeted duties on dozens of countries. That could jack up the price for U.S. shoppers of everything from cannabis to running shoes to Apple's iPhone. A high-end iPhone could cost nearly $2,300 if Apple passes the costs on to consumers, based on projections from Rosenblatt Securities. Canadian Prime Minister Mark Carney said the U.S. had abandoned its historic role as a champion of international economic cooperation. China vowed retaliation for Trump's 54% tariffs on imports from the world's No. 2 economy, as did the European Union, which faces a 20% duty.  The tariffs "clearly represent a significant risk to the global outlook at a time of sluggish growth," said IMF Managing Director Kristalina Georgieva, calling on Washington to work to resolve trade tensions with its partners and reduce uncertainty. American companies with significant overseas production took a hit. Nike shares lost 14% and Apple fell 9%.

"The tariff plan does not appear to be well thought-out. Trade negotiations are a highly technical discipline, and in our view these proposals do not offer a serious basis for negotiations with any country," said James Lucier, founding partner at Capital Alpha.
Economists say the tariffs could reignite inflation, raise the risk of a U.S. recession and boost costs for the average U.S. family by thousands of dollars.

Intel and TSMC have tentativiely agreed to a "joint venture" with Intel Foundry, which would allow TSMC to manage Intel's US facilities. Here's why its a bad idea for Intel and the US: 

1./ TSMC and Intel are two VERY different companies with unique management, workforce, and technology roadmap planning, meaning that fundamentals cannot be merged. And, given that TSMCs ambitions in the US, partnering up with Intel means that the company is having a pretty difficult task on its hands. The only way this could succeed is if TSMC could obtain "complete" management control, but this will be a process that will involve years if not decades. So any alleged benefits by such a JV would still be a long way off

2./ If you look at Intel Foundry's recent progress, the division has managed to find a hold on the market, having recently announced "Risk Production" of its highly anticipated 18A process, an achievement that could mark the firm's economic turnover.

3./ Supposedly part of Intel's new CEO Lip-Bu Tan, core focus will be to "build for clients," with foundry services being a priority, so a TSMC collaboration doesn't look correct or in keeping with Fiduciary Responsibility of "duty of care", "duty of obedianc", and "duty of loyalty" here as one would reasonably expect. 

Samsung has turned to Chinese technology groups to prop up its ailing semiconductor division, as it struggles to secure big US customers despite investing tens of billions of dollars in its American manufacturing facilities. The South Korean electronics group revealed last month that the value of its exports to China jumped 54% between 2023 and 2024, as Chinese companies rush to secure stockpiles of advanced AI ICs in the face of increasingly restrictive US export controls. In one previously unreported deal, Samsung in 2024 sold +3 years’ supply of logic ICs—a key component in manufacturing AI chips—to Kunlun, the semiconductor design subsidiary of Chinese tech group Baidu. The increasing importance of its China sales to Samsung comes as it navigates growing trade tensions between US and China over the development of sensitive technologies.

Samsung announced in 2024 that it was making a $40 Billion investment in expanding its advanced IC manufacturing and packaging facilities in Texas, boosted by up to $6.4 Billion in federal subsidies.  But Samsung’s contract chipmaking business has struggled to secure big US customers, bleeding market share to TSMC, which is investing “at least” $100 billion in IC fabs in Arizona.

“Samsung and China need each other,” said CW Chung, joint head of Apac equity research at Nomura. “Chinese customers have become more important for Samsung, but it won’t be easy to do business together. Samsung has also fallen behind local rival SK Hynix in the booming market for “high bandwidth memory,” another crucial component in AI chips. As the leading supplier of HBMs for use by Nvidia, SK Hynix’s quarterly operating profit last year surpassed that of Samsung for the first time in the two companies’ history. “Chinese companies don’t even have a chance to buy SK Hynix’s HBM because the supply is all bought out by the leading AI chip producers like Nvidia, AMD, Intel, and Broadcom,” said Jimmy Goodrich, senior adviser for technology analysis to the Rand Corporation research institute. “What Samsung is producing are the scraps that are inferior but still good enough for the Chinese, as there is no local HBM substitute yet.” Samsung is the “biggest supplier of HBM into China,” which is used in Huawei’s Ascend 910 series of AI chips.

Samsung's contract chipmaking business also partnered with Kunlun to produce the Core P800, an AI chip released in February that also incorporates Samsung HBM. A person familiar with Samsung’s thinking said it had hoped to work with Kunlun to produce an even more advanced AI chip, but the project had been put on hold because of new US restrictions that came into force in January.

The curbs, which cap the permitted performance of AI chips produced for Chinese customers by foreign foundries, were tightened after TSMC acknowledged last year it had mistakenly assembled AI chips for shell companies acting on Huawei’s behalf.

“Our business with Baidu has become uncertain since stronger US export controls took effect in January,” said a person close to Samsung, who added it was seeking more leeway from US authorities.  But he noted the restrictions meant that Chinese companies “will take anything they can get, either in terms of HBM or logic foundry capabilities, as Chinese indigenous players are not yet as competitive.” “Samsung producing what may be a very competitive chip for Baidu raises questions as to whether the performance thresholds of US restrictions need to be adjusted—the longer US officials wait, the more of these chips are going to be produced for China.” 

A 43-year-old Russian engineer is accused of secretly supplying sensitive technical information from ASML, NXP, and TSMC to Russia, allegedly to assist in building a 28 nm-capable fab there. The engineer, identified in court documents only as German A., earned about €40,000 and now faces 18-32 months in prison. Though German A. alone could not steal full designs for a semiconductor.  German A. is accused of supplying Russia with confidential technical materials from ASML, GlobalFoundries, NXP, TSMC, and GlobalFoundries, including semiconductor production manuals and various chipmaking machines. The investigators reportedly found that he obtained 105 internal documents from ASML and 88 files related to TSMC. The materials did not contain complete blueprints for building wafer fabrication equipment or something more significant (e.g., a fab itself or how to design a process technology). Still, they were labeled confidential and could support the setup of a basic semiconductor line capable of producing chips at 28- nm-class process technology, which is good enough for military applications. Investigators believe he shared this data via cloud storage and messaging apps and handed over a USB stick in Moscow, allegedly earning around €40,000 in the process.  Both ASML and NXP experienced breaches involving unauthorized access in the past. In late 2023, it was revealed that a cyber group linked to China had been covertly operating within NXP's systems for an extended period. ASML also battles frequent cyberattacks and insider threats: in early 2022, a former Chinese employee stole confidential data. Although that employee, just like German A., lacked access to complete designs needed to construct a fab tool or equip a fab, a broader network of similar operatives could realistically piece together enough to boost the semiconductor industries of China and Russia.

RiVAI Technologies has launched the Lingyu CPU, China’s 1st domestically designed HPC (High-Performance) RISC-V server processor, reflecting the country's ongoing push for greater self-sufficiency in semiconductor development. The Lingyu CPU adopts a one-core, dual architecture approach, integrating 32 general-purpose computing cores (CPU) alongside 8 specialized intelligent computing cores (LPU). The configuration efficiently handles tasks such as inference for open-source large language models. The architecture aims to balance processing power and energy efficiency, thereby lowering the total cost of ownership (TCO).  The push for RISC-V adoption comes in response to ongoing trade tensions and sanctions that have limited China’s access to advanced foreign-made chips. To accelerate this transition, the Chinese government provides policy support, funding, and incentives for companies working on RISC-V technology. Major domestic tech firms, including Alibaba and Tencent, have already started developing RISC-V-based solutions, while state-backed research institutions are working on software optimization for the architecture. This shift could help China build a more self-sufficient semiconductor industry, reducing its dependence on Western technologies. However, challenges remain, including software compatibility and ecosystem development, which will determine the long-term viability of RISC-V as a mainstream alternative to x86 and Arm processors.

U.S. trading partners threatened to ratchet up a trade war with the United States on Thursday as President Donald Trump's sweeping tariffs ignited fears of steep price hikes in the world's largest consumer market. Stocks suffered a global meltdown, as analysts warned the tariffs could upend global supply chains and hurt corporate profits. The Dow fell nearly 4%, its biggest one-day loss in percentage terms since June 2020. The S&P 500 lost nearly 5% and the tech-heavy Nasdaq declined nearly 6%, its worst day in percentage terms since the pandemic era of March 2020. The penalties announced by Trump on Wednesday triggered a plunge in world markets and drew condemnation from other leaders reckoning with the end of a decades-long era of trade liberalization. The U.S. tariffs would amount to the highest trade barriers in more than a century: a 10% baseline tariff on all imports and higher targeted duties on some of the country's biggest trading partners. That could jack up the price of everything from cannabis to running shoes to Apple's iPhone for U.S. shoppers. China vowed retaliation for Trump's 54% tariffs on imports from the world's No. 2 economy, as did the EU, which faces a 20% duty. French President Emmanuel Macron called for EU countries to suspend investment in the US. Other trading partners, including South Korea, Mexico and India, said they would hold off for now as they seek concessions." The consequences will be dire for millions of people around the globe," EU chief Ursula von der Leyen said. Trump has slapped a 24% tariff on Japan and a 25% tariff on South Korea, both home to major U.S. military bases, and htting Taiwan with a 32% tariff as the island faces increased military pressure from China. In Europe, Trump has already upset NATO allies with demands for higher defense spending and potential concessions to Russia in its war in Ukraine. Canada and Mexico, the largest U.S. trading partners, were not hit with targeted tariffs on Wednesday, but they already face 25% tariffs on many goods and now face a separate set of tariffs on auto imports.

Arm Holdings expects its share of the global market for data center central processing units to surge to 50% by EOY 2025, up from about 15% in 2024 with gains driven by the boom in AI, a senior executive said. It struggled to make headway in the lucrative data center market for nearly two decades as switching over from the once-dominant x86 chips made by Intel and AMD means clients have to rewrite software as well as change up parts of hardware. "We've gotten to the point where software is actually being developed for Arm first and foremost,". Amazon has designed in-house Data Center CPUs with Arm tech that accounted for +1/2 of the capacity for chips it added over the last two years. Alphabet's Google and Microsoft have also made Arm-based data center chips, though their efforts are more recent than Amazon's.  Arm's CPUs are often used as a "host" chip inside of an AI computing system and act as a kind of traffic controller for other AI chips. Nvidia, for example, uses an Arm-based chip called Grace in some of its advanced AI systems which contain two of its Blackwell chips. Arm's tech in many cases offers lower power consumption than rival processors made by Intel and AMD. As AI data centers use huge amounts of electricity, Arm's chips have become increasingly popular among cloud computing companies. Awad added that data center chips often use more of Arm's intellectual property (IP) and the company typically receives "a lot higher aggregate royalty rate" than chips for less complex devices. UK-headquartered Arm, which is 90% owned by Japan's SoftBank Group, does not make chips itself but sells the fundamental building blocks and other intellectual property to cloud computing companies and firms like Apple and Nvidia to use to design chips for laptops, smartphones and data center processors. Arm generates revenue by billing companies for a license to use its tech and collects royalty payments for each chip sold.

The Hsinchu and Kaohsiung plants in Taiwan serve as the primary facilities for 2-nm wafer production, with a new report stating that full-scale manufacturing of this technology is expected to commence later this year. Before this phase, TSMC had achieved an impressive 60% yield during the trial production. With both facilities operational, the semiconductor giant could churn out up to 50,000 monthly wafers, with the maximum capacity of 80,000 units. The demand for 2-nm chips is said to be higher than 3nm, and according to fresh estimates, TSMC can generate sales of $30.1 billion for both Q3 and Q4 2025.  Pilot production on 2nm technology at the Hsinchu plant has exceeded exceptations, allowing TSMC to commence operations a quarter ahead of schedule  According to Economic News Daily, both the Hsinchu and Kaohsiung will handle the imminent demand, with Dan Nystedt mentioning on X that each of these plants can bring in sales of ~$30.1 Billion for both Q3 and Q4 2025.   Much of this revenue will be thanks to Apple, which is said to be preparing the A20 for the iPhone 18 launch that is scheduled to happen in H2 of 2026. Qualcomm also intends to keep pace with Apple, as it is rumored to unveil not one, but two SoCs that will be mass produced on TSMC’s 2nm process, of which one of them will likely be the Snapdragon 8 Elite Gen 2. Perhaps the only competition TSMC has at this time is Samsung, which was previously reported to have achieved a 30 percent yield on its 2nm GAA process during the trial production run of the company’s Exynos 2600. Historically, TSMC might have been late in gravitating to the newer lithography, but it has always stood tall when maintaining consistent wafer output.

Intel has reportedly canceled its high-end Xe2-based Arc Battlemage "BMG-G31" GPUs which were going to be used in gaming graphics cards, has decided to NOT Compete In The High-End Gaming Segment  Intel has stated its commitment towards the discrete GPU segment, but will continue to make "Strategic Investments" within the platform.  Based on the latest report by reliable insider and leaker, @Jaykihn0, it looks like the company has significantly altered its discrete GPU plans. It is said that the high-end Battlemage BMG-G31 GPU is more or less dead and that has been the case since Q3 of 2024. The BMG-G31 GPU die was supposed to be bigger than the G21 featured on the B580 and B570 graphics cards. It was reportedly going to feature around 24-32 Xe2 cores with a 256-bit memory bus and 16 GB of GDDR6 memory. It is also mentioned that there's currently no update on the Celestial "Xe3" discrete GPU lineup. The Xe3 architecture will be deployed in the next-gen Panther Lake CPUs that are being fabricated on the Intel 18A process node, but that's an integrated solution. Whether we will get a Celestial "Xe3" GPU for discrete use remains to be seen. With the Arc B580 and Arc B570, Intel has provided strong value in a segment where NVIDIA and AMD currently have no new options. Battlemage also proves itself to be an impressive solution for integrated graphics and competes well against AMD's latest RDNA 3.5 offerings. Although we would love to see a high-end Battlemage graphics card, it looks like the company is going to focus its GPU efforts on the integrated sides a bit more but will continue to provide gamers with some decent value-oriented products, such as its recent launches.

TSMC's former co-Chief Operating Officer, Chiang Shang-yi, didn't hold back when talking about Intel's chip manufacturing woes at an event in Taiwan. Chiang advised Intel to move into mature chip manufacturing processes to win, as it was too far behind TSMC to catch up. Chiang added, while Intel was formerly a "King" of the IC industry, it was now a "Nobody."   With Intel's new CEO, Lip-Bu Tan, settling into his role, the firm has 2 primary objectives. The 1st is to establish regular production of its 18A chip manufacturing node, and the 2nd is to set up a robust foundry business. Both of these are aimed at competing with TSMC. Should Intel execute with the former, then it will have achieved manufacturing node parity with TSMC, which also plans to place the comparable 2-nm process into HVM in 2025.  While Intel might try at another turnaround attempt, TSMC's Shang-yi believes that Intel would be better suited to focus on more mature manufacturing process nodes, instead of competing with TSMC on the high-end nodes, Intel should merge with a company that does not produce leading-edge chips but instead churns out mature process chips in large volumes.  As for TSMC, he believes that the firm's biggest advantage lies in its 100s of customers. To serve the voluminous customer base, TSMC has established high output and quick manufacturing speeds, which are its key competitive advantage. Shangyi said TSMC's global dominance and success over UMC (another Taiwanese contract manufacturer) was the firm's focus on R&D. While UMC worked with IBM for research purposes, TSMC self-developed the capabilities. He recalled fondly how hard he worked to overcome Intel's dominance and is happy after witnessing TSMC doing so.

A significant concern has arisen following Lip-Bu Tan's appointment:

Extensive Investments in China: Through his venture capital firm Walden International and other holding companies (Sakarya Limited, Seine Limited), Tan controls +40 Chinese companies and holds minority stakes in +600 others, with these stakes valued at potentially +$200 million.  -- Critically, investments have been made in, or alongside, Chinese companies linked to the People's Liberation Army (PLA), Chinese state-owned groups, or government funds. These Concerns center on investments in firms like SMIC (China's top foundry, Tan was an early investor and board member until 2018; Walden reportedly exited in 2021 after SMIC was sanctioned by the US for military ties), Dapu Technologies (identified as a PLA contractor), HAI Robotics (reportedly bids for PLA contracts and works with surveillance firms), Intellifusion (a surveillance tech firm on a US trade blacklist for alleged human rights abuses), QST Group (whose sensors were reportedly found in Russian military drones), and Wuxi Xinxiang (supplier to YMTC, which is on US blacklists).  
 Given Intel's vital role as a US defense contractor and a key player in national technology infrastructure, Tan's deep financial integration with entities in a strategic competitor nation, particularly those with military or state ties, raises serious national security alarms and potential conflict of interest issues. Critics question his suitability to lead Intel under these circumstances. The past investment in SMIC is also seen by some as problematic for the CEO of a company trying to build its own foundry business.  -- These investment patterns have drawn scrutiny from the U.S. Congress (House Select Committee on the CCP).  While holding these investments is legal unless a company is on a specific U.S. restriction list, ethical concerns are raised regarding ties to surveillance firms or military suppliers. Intel has stated Tan completed required SEC disclosures regarding potential conflicts. There's discussion about whether divestment from potentially conflicting holdings should be required.

The paper examines the broad range of methods, tools and strategies available to designers and attempts to distill the best of each in a bid to generate a coherent, 'systematic creative design' philosophy. Although using the Soviet-originated Theory of Inventive problem Solving, TRIZ as its foundation, the proposed design method also encompasses elements of, amongst others, QFD, Design for X, Value Engineering, Axiomatic Design and Robust Design. The paper describes the ongoing process of integrating these methods and reviews their deployment on a broad spectrum of real engineering design case studies.

Some Intel investors and Wall Street analysts remain pessimistic about the prospects of a TSMC and Intel JV ending in any kind of a positive outcome.  Intel and TSMC have reached an understanding to constitute a JV that would then manage Intel's US-based fabs. The arrangement would also rope in other IC designers such as Qualcomm, NVIDIA, and Apple. TSMC would reportedly retain a 20% stake in the JV, financed via its in-kind contribution of technology and expertise.  This arrangement, however, raises quite a few questions. After all, Intel and TSMC have completely different chip fabrication processes that are barely compatible with each other, if at all, and it is for this reason that Citi has now cast serious doubts on the viability of the proposed JV:  "We do not believe TSMC operating/forming a JV with Intel would work given differences in manufacturing and operations." Citi then goes on to "question the wisdom of fabless companies investing in this JV," noting: "We believe Intel foundry has proven over the] years it cannot compete with TSMC, and forcing a company to use vastly inferior manufacturing would destroy the shareholder value of a fabless company such as Qualcomm or Broadcom."  Instead, Citi thinks Intel should exit the chip fabrication business entirely: "Given the highly unlikely chance of Intel merchant foundry succeeding and subsequent drag on cash flow, we continue to believe Intel would be best served by exiting the merchant foundry business and focusing on its core CPU business."

BofA's Vivek Arya also recently denigrated this proposal, noting that splitting up Intel "could be time-consuming and complicated given:

1) Extensive approval requirements from global (China) regulators due to Intel's dominant ~70% share of PC/server CPU

2) Mismatch between Intel and TSMC’s manufacturing processes,

3) TSMC’s ongoing fab expansion plans in Arizona with the ability to serve AI customers,

4) Broadcom's existing low-levered but high-debt ($ 58 billion net debt) balance sheet, and

5) Constraints of Intel's CHIPS Act funding (design needs to own 50 %+ of manufacturing) and ROI requirements from co-investment partners ..."

Private equity firms prioritize profits over people, buying and flipping businesses across industries, which leads to degraded services, inflated costs, and widespread corporate control that undermines competition and societal well-being. Private equity is identified as a major yet overlooked factor contributing to economic struggles. Private equity firms manage investments from wealthy individuals, often using leverage to acquire businesses. Private equity firms prioritize profit over the well-being of the businesses they acquire.

The Dimensity 9400 is substantial proof that MediaTek is no longer a company that is sidelined to deliver entry-level and mid-ranged chipsets, but that does not explain why the company has not introduced a Snapdragon X Elite competitor, or, at the very least, a Snapdragon X Plus one. Turns out it has, but this SoC will be found in a completely different set of machines for whatever reason. The Kompanio Ultra 910 is the newest release from the Taiwanese firm, and looking at its specifications, we are definitely excited about what comes next. -- Chromebooks featuring the Kompanio Ultra 910 will be available in the coming months, but the latest chipset release sounds like a missed opportunity for MediaTek  -- The Kompanio Ultra 910 is fabricated with the best available technology, with MediaTek stating that the silicon is mass produced on TSMC’s 2nd-gen 3-nm process, the same one utilized for Apple’s M4, M4 Pro, and M4 Max. The Kompanio Ultra 910 sports no low-power cores, with its 8-core configuration sporting an Arm Cortex-X925 operating at up to 3.62GHz, a Cortex-X4 and Cortex-A720, the chipset can achieve a high bandwidth as it supports LPDDR5X RAM up to 8,533MHz.

The new chipset also flaunts impressive AI capabilities thanks to MediaTek’s 8th-gen NPU, which delivers up to 50TOPS AI performance and faster LLM speculative speed support. Even though the Chromebooks outfitted with the Kompanio Ultra 910 are not intended for gaming, MediaTek claims that the 11 GPU cores belonging to the Immortalis-G925 deliver ‘PC-grade raytracing capabilities’. The Kompanio Ultra 910 can also support up to two 4K external monitors to boost productivity, with the inclusion of Wi-Fi 7 up to 7.3Gbps and Bluetooth 6.0. It is surprising that MediaTek’s latest silicon will only be found in Chromebooks later in 2025. It is possible that Qualcomm’s and Microsoft’s exclusive agreement of only using the former’s Snapdragon range of chipsets prevents competitors like MediaTek from stepping foot into this realm. MediaTek has an ace up its sleeve by partnering with NVIDIA to debut an AI PC in H2 2025, with the impending release likely stirring up the competition to newfound levels.

Well, it seems like the DRAM industry is expected to witness uncertainty once again, as Samsung is reportedly planning to implement a global price rise. This surely isn't a good time for consumers at all, since, apart from all the uncertainty in the market associated with Trump tariffs, companies are reportedly trying to combat them by "front-running" the announcements and importing products in huge volumes to avoid additional taxation. Similarly, according to the Korean media outlet MK, the Korean giant plans to increase DRAM and NAND prices by up to 3%-5%, keeping in light the demand from the semiconductor industry and how competitors like Micron are increasing prices as well. Based on the data from DRAMeXchange, DDR5 memory module pricing has risen by up to 12%, and the demand was mainly from HPC servers, while NAND pricing saw a 9% rise. This has shown that demand is definitely there, but it isn't from the consumer PC segment but rather from the AI market, which is why prices have seen such a rise. With companies like Samsung and Micron announcing a price hike, this means that prices would soar further ahead, which isn't good for the consumer markets. Consumer RAM pricing could rise notably in the upcoming months, and it won't be wrong to say that now might be the best time to get your DDR5 modules, given that with Trump tariffs and trade uncertainities, the consumer PC markets will see a tremendous rise in pricing.

Microsoft has pulled back on Data Center projects around the world, Bloomberg reports, suggesting that the company is wary of expanding its cloud computing infrastructure too rapidly. Microsoft has halted talks for or delayed development sites of data centers in the U.K., Australia, North Dakota, Wisconsin, and Illinois, per Bloomberg. A spokesperson told the publication that Microsoft makes its plans years in advance and that the changes demonstrate “the flexibility of [its] strategy.” As recently as February, Microsoft reiterated earlier plans to allocate more than $80 billion of its cash to capital expenditures in 2025, primarily AI data centers. As Bloomberg points out in its piece, it’s hard to know how much of the company’s recent pullback reflects its expectations of diminished demand versus temporary construction challenges, such as shortages of power and building materials. Microsoft previously said that it would shift its Data Center expansion focus for 2025 from new construction to fitting existing facilities with servers and other computing equipment.

Arm Holdings just made a quiet but telling move in the AI chip war by pursuing a potential acquisition of UK-based Alphawave to lock in next-gen tech, then backing out. The tech in question, a high-speed data transfer solution called serializer-deserializer (SerDes), is the unsung hero behind how ICs talk to other essential ICs for training and running AI models at scale. Alphawave shares popped nearly 45% on the news, their best day in +3 years. While the deal fizzled out, the message was clear: Arm is thinking bigger than just licensing IP.

Why does it matter? Because SerDes isn't just another piece of silicon, it's the plumbing for tomorrow's AI infrastructure, and right now, Broadcom, Marvell, and even Nvidia are way ahead. Arm, backed by SoftBank and traditionally a behind-the-scenes IP powerhouse, doesn't currently offer SerDes at the bleeding edge. There's more at play here, as Alphawave has ties to a blacklisted Chinese firm, that may have thrown a wrench into the talks. The strategic takeaway is that  Arm wants in on the $60 Billion custom AI IC market projected by 2028. And SerDes is the toll gate. If they can't buy it, they'll have to build it and fast. Because as demand for HPC (High-Performance Compute) AI chips surges, whoever controls the chip-to-chip highway will end up owning the map. Building it from scratch? Insiders say it takes a specialized team and at least 2 years. Which explains why Arm has quietly started hiring chip designers and floating internal memos about launching its own processors.

Suraj Gajendra, Vice President of Automotive Products and Software Solutions at Arm focuses on Arm's latest innovations in the automotive sector, including the introduction of Kleidi AI, a suite of software libraries for optimizing AI applications on Arm CPUs, and the critical role of chiplets and standardization in enabling scalable compute architectures. Suraj Gajendra's insights reflect a forward-thinking approach to automotive technology, emphasizing the importance of AI, modular hardware, standardization, and collaboration to shape the future of automotive computing. Understanding these principles is essential for anyone involved in the automotive industry. Sharing insights on Software-Defined Vehicles (SDVs), the importance of cloud-to-car infrastructure, and the need for robust security strategies to ensure safe and resilient automotive solutions. He emphasizes the significance of AI in enhancing vehicle performance and user experience, and introduces Cy AI, which focuses on optimizing automotive applications. He 

advocates for the use of chiplets to create a modular and scalable hardware architecture, and highlighted how chiplets can reduce costs and improve flexibility across different vehicle models. Suraj stresses the importance of standardization in chip design and software development to ensure compatibility and interoperability, encouraging a collaborative approach among industry players to establish common standards.  He explains the evolution of SDVs and their reliance on software for functionality and updates with an emphasis on the importance of safety, security, and data protection in software development. 

China's Huawei is expected to claim triumph over U.S. sanctions at its upcoming annual results, bolstered by its software push, progress in chips and booming smart-driving technology business that has helped it move out of "survival mode". Hauwei is set to confirm that it took in $118 billion in revenues in 2024, just shy of its 2020 peak of 891 billion yuan, before chip stockpiles dwindled and U.S. restrictions cut consumer business revenues in 1/2. In the wake of U.S. sanctions, Huawei moved into exploring areas such as building 5G infrastructure for mines and supplying energy storage systems to Data Centers. Cut off from Google's Android and Oracle, it built its own operating system HarmonyOS, which it says is running on over a billion devices, as well as an internal software management system it calls 'MetaERP' Banned from using U.S. semiconductor technology, it has created its own advanced chips including ones that compete with top artificial intelligence chipmaker Nvidia's products. Hauwei has also become a prominent supplier of advanced autonomous driving technology, working with state-owned automakers to revive themselves as viable electric vehicle makers. Huawei has worked with Dongfeng Motor-backed Seres to sell Aito-branded cars, with sales more than tripling last year. Its best-selling models M7 and M9 are equipped with Huawei's advanced driver assistance systems and sold in Huawei's showrooms nationwide. There are similar projects with Chery, BAIC, JAC Group and SAIC Group. Going forward, the company has said it wants to integrate artificial intelligence into its industrial communications services and to build out its software systems on connected devices, according to state media. Huawei has also signaled it intends to compete more aggressively in overseas markets for its smartphones, having launched its foldable Mate XT smartphone in Malaysia in February in a glitzy event.
Without full access to Android it is unlikely to regain its former position in Western consumer markets, though its data infrastructure presence has grown in areas such as the Middle East, Triolo said.
"Huawei's international presence will be more of a patchwork affair, but in some areas, like an alternative AI stack, it could eventually dominate in key markets."

ASML's CEO, Christophe Fouquet, is concerned with geopolitical pressures and the lack of strong, independent support from EU policymakers for his strategically important company, key concerns are: -- US-led export restrictions to China, banning selling advanced (EUV) and some older (DUV) machines to Chinese companies, driven by US efforts to curb China's AI ambitions. The likelihood of even tighter controls under a 2nd Trump administration adds significant uncertainty and makes long-term planning difficult.
-- Even if ASML were to consider leaving the EU, it would still be subject to US export controls due to its reliance on US-based Cymer technology, highlighting the limitations of its strategic autonomy.
-- Fouquet's frustrated by EU governments, particularly the Dutch government, readily following the US lead on export bans. Believing Europe needs to define its own interests and not be dictated to by others. and recognize ASML's strategic importance.
-- Fouquet emphasizes that the semiconductor industry's success is built on a delicate "chain of trust," deep specialization, and collaboration. He fears that "overboard" government decisions could break this chain, jeopardizing the technological advancements we currently take for granted.