March 26, 2026
Japan consolidates. Tesla builds a fab. The Gulf buys compute access. Zoho owns its servers. The pattern is the same everywhere.
Over the past few weeks I have been writing about the semiconductor consolidation happening in Japan, first the Denso bid for Rohm, and today the announcement that Mitsubishi Electric, Rohm, and Toshiba will begin merger talks to form the world's second-largest power chip group. These are not isolated events. They are part of a race that is now happening on every continent, across every category of technology company, and at every level of the stack. The insight worth paying attention to is not any single deal. It is the pattern underneath all of them.
Chips are present in more of daily life than most people realize. Modern electric vehicles contain between 2,000 and 3,000 chips. Headphones, home appliances, traffic systems, industrial equipment, hospital devices, and the data centers running AI inference all run on semiconductors. The pandemic supply chain disruption made this visible in a jarring way, particularly in the automotive sector where production lines stopped because a $5 chip was unavailable. That moment changed how executives, governments, and technology companies think about semiconductor access. It stopped being a procurement question and became a continuity question.
The AI Demand Curve Made It Urgent
What the pandemic exposed, the AI buildout accelerated. Every hyperscaler, every enterprise deploying AI at scale, and every company building next-generation products now faces the same constraint: chip capacity is fully booked for years. Google, Amazon, and Meta are not just buying chips. They are designing their own. Google has its Tensor Processing Units. Amazon has Trainium and Inferentia. Meta has its Meta Training and Inference Accelerator. The largest consumers of compute concluded that waiting in line for someone else's silicon was an unacceptable strategic position.
Tesla and SpaceX made the same calculation this month. Elon Musk announced Terafab on March 21, a joint venture between Tesla, SpaceX, and xAI targeting a vertically integrated chip fabrication facility in Austin, Texas. The stated goal is one terawatt of annual AI compute capacity. The explicit rationale was direct: current chip manufacturers produce only about 2 percent of what Tesla and SpaceX will need for their future compute requirements. The framing Musk used is worth repeating, "We either build the Terafab or we don't have the chips." That is not a technology ambition statement. It is a supply chain risk statement.
Japan's Answer: Scale Through Consolidation
Japan cannot write a $25 billion check from a single company the way a Musk enterprise might attempt. Its answer is consolidation. Mitsubishi Electric, Rohm, and Toshiba merging their power semiconductor businesses is the mechanism that makes continued investment possible at the scale required. No single Japanese manufacturer has the balance sheet to fund next-generation silicon carbide and gallium nitride development cycles while also competing with Infineon on cost. Together, they can. This is Japan's way of building the floor, not by one company absorbing the others, but by creating a shared platform for national competitiveness.
The technology at stake matters for the broader AI story. Silicon carbide power semiconductors are the components that make energy conversion efficient enough to run AI data centers without straining the power grid. They are also what makes electric vehicle inverters viable at scale. Gallium nitride is the next layer, handling high-frequency power delivery in compact data center infrastructure. Whoever controls these materials and the manufacturing processes behind them holds significant leverage over the physical infrastructure of the AI era.
The Middle East: Buying Access First, Then Building
The Gulf states are playing a different version of the same game. Saudi Arabia and the UAE cannot build advanced fabrication plants overnight. Their strategy is sequenced. First, secure chip access. In November 2025, the United States authorized the export of advanced Blackwell chips to HUMAIN in Saudi Arabia and G42 in the UAE, up to 35,000 chips each, as part of what the US government is calling compute diplomacy. The Stargate UAE consortium, anchored by G42 in Abu Dhabi, broke ground on a one-gigawatt AI compute cluster. The UAE's Mubadala Investment Company has a stake in GlobalFoundries, giving it a position in actual fabrication rather than just procurement.
Saudi Arabia is building toward design capability. The National Semiconductor Hub, backed by a fund exceeding $266 million, targets at least 50 chip design companies in the kingdom by 2030 and a training pipeline of over 5,000 semiconductor engineers. This is not a fast path to a fab. It is a deliberate effort to own intellectual property and talent so that the region is not permanently dependent on whoever holds the manufacturing. Egypt has a similar instinct, focusing on chip design rather than fabrication, building on an existing base of electrical engineering talent and attracting design centers from global manufacturers.
What Zoho Understood Early
In February I wrote about Zoho's infrastructure strategy coming out of ZohoDay26. What Zoho's leadership articulated was a version of this same instinct applied one layer up. Rather than depending on hyperscaler cloud margins and token pricing they cannot control, Zoho is building custom server hardware, designing its own motherboards and firmware, and running its own data centers. The goal is to own the layers where AI costs originate, so that what gets passed to customers is a predictable price, not a variable tied to someone else's infrastructure decisions. Zoho's chief information officer described an expectation that a significant amount of processing will eventually move on-premises, running on compute inside the business facility itself.
This is the enterprise-level version of what Japan, Tesla, and the Gulf states are doing at the national and mega-corporate level. Own what you depend on. The layer each actor is targeting differs. The logic is identical.
The Branding Gap in the Middle of All This
There is a consumer perception problem worth naming. When most people hear "chip company," they think Nvidia. Possibly Intel or AMD. That is a marketing and branding outcome, not a technology reality. Nvidia has done exceptional work making its brand synonymous with AI hardware in the public mind. But the semiconductor ecosystem that runs the world is far larger and largely invisible: power chips in cars, memory in data centers, sensors in appliances, modem chips in headphones, controller chips in industrial equipment. The companies building those components, and the nations racing to control them, are not household names. That gap between public perception and strategic reality is part of why the consolidation happening in Japan and the investment happening in the Gulf has received less attention than it deserves.
The Question for Technology Leaders
The viability question this raises for CIOs and CTOs is not which company wins any individual deal. It is whether your organization's technology roadmap accounts for a world where semiconductor supply is increasingly allocated through strategic relationships, national policy, and vertical integration rather than open market procurement. The companies and countries that recognized this dependency early are building positions now. Those that still treat chip access as a commodity purchasing decision will find their options narrowing as consolidation reduces the number of suppliers and demand continues to outpace capacity.
The hardware floor sets the software ceiling. That argument has not changed. What has changed is how many actors, in how many places, are now racing to control it.
Related reading on shashi.co
Bellamkonda, Shashi. "Why Your Software Strategy Depends on This $8B Chip Deal." Shashi.co, 6 Mar. 2026, shashi.co/2026/03/why-your-software-strategy-depends-on.html.
Bellamkonda, Shashi. "Japan's Power Chip Consolidation Just Got Bigger — and More Complicated." Shashi.co, 26 Mar. 2026, shashi.co/2026/03/japans-power-chip-consolidation-just-got.html.
Bellamkonda, Shashi. "When the Chip Race Leaves Earth." Shashi.co, 21 Mar. 2026, shashi.co/2026/03/when-chip-race-leaves-earth.html.
Bellamkonda, Shashi. "Building From the Ground Up: Zoho's Bet Against AI Infrastructure Dependency." Shashi.co, 21 Feb. 2026, shashi.co/2026/02/building-from-ground-up-zohos-bet.html.
Bellamkonda, Shashi. "From Trunk Calls to 5G Radio: India's Long Road to Telecom Self-Reliance." Shashi.co, Mar. 2026, shashi.co/2026/03/from-trunk-calls-to-5g-radio-indias.html.
Bellamkonda, Shashi. "The Layer Nobody Talks About: Lenovo's GTC Announcements and the AI Deployment Problem." Shashi.co, Mar. 2026, shashi.co/2026/03/the-layer-nobody-talks-about-lenovos.html.
Bellamkonda, Shashi. "Lumen Technologies: The Rise of the Shadow Hyperscaler." Shashi.co, Feb. 2026, shashi.co/2026/02/lumen-technologies-rise-of-shadow.html.
Sources
Shimizu, Ritsuko, Kantaro Komiya, and Daniel Leussink. "Rohm, Toshiba, Mitsubishi Electric to Begin Power Chip Integration Talks, Nikkei Says." Reuters, 26 Mar. 2026, finance.yahoo.com/sectors/technology/articles/rohm-toshiba-mitsubishi-elec-begin-085637172.html.
Musk, Elon. "Announcing TERAFAB: The Next Step Towards Becoming a Galactic Civilization." Tesla, 22 Mar. 2026, tesla.com.
"Elon Musk Announces Terafab Project He Claims Will Be the 'Largest Chip Manufacturing Facility Ever.'" Engadget, 22 Mar. 2026, engadget.com/science/elon-musk-announces-terafab-project-171718545.html.
"US Authorizes Chips for the UAE, Saudi Arabia." Middle East Institute, 17 Feb. 2026, mei.edu/policymemo/us-authorizes-chips-for-the-uae-saudi-arabia-2/.
"Saudi Arabia Boosts Chip Ambitions, Launches Semiconductor Hub." MIT Sloan Management Review Middle East, 10 Jun. 2024, mitsloanme.com/article/saudi-arabia-boosts-chip-ambitions-launches-semiconductor-hub/.
"Why Local Manufacturing of Semiconductors Is a Big Deal for the Middle East's Smart Economy." Fast Company Middle East, 20 May 2024, fastcompanyme.com/impact/why-local-manufacturing-of-semiconductors-is-a-big-deal.
Ramamoorthy, Ram Prakash. "Foundations that Scale." ZohoDay25 Presentation, 2026, zoho.com.
