MediaTek AI Data-Center Chip Expansion: Inside the $5B Bet and the 20% Phone Slump
MediaTek, already the world's largest supplier of smartphone chips, has launched an AI data-center chip expansion backed by a discretionary financing budget of $5 billion, approved on the same day it reported a 20% year-over-year fall in mobile-chip revenue. Taiwan's largest chip designer said on July 31 that the framework will fund long-term growth, including custom AI accelerators, or ASICs, built for major cloud providers. The plan pairs an aggressive move into data-center silicon with a smartphone business that shrank sharply in the second quarter.
Chief Executive Rick Tsai framed the budget as a flexible, discretionary framework that gives the company optionality to move quickly on data-center opportunities as they arise. Because no single program is tied to the full amount, the approval signals balance-sheet capacity for large outlays without committing MediaTek to one spending path. The board authorization also works as a standing facility that management can draw on as programs mature, rather than a single upfront capital expenditure, and the description of the plan as an optionality play points the same way.
The company's stated ambitions in custom AI silicon go well beyond its current revenue base. MediaTek raised its estimate of the 2027 serviceable addressable market for custom AI chips to $80 billion, a point estimate at the top of its previous $70 billion to $80 billion range, and lifted its target share of that market to 15% to 20% from 10% to 15%. At the lower end of the new target, that implies about $12 billion in annual custom-chip revenue by 2027; at the upper end, $16 billion, a six- to eight-fold increase over the more than $2 billion the company expects from the data-center business this year. The estimate now sits at the top of the prior range, a sign that management expects hyperscaler budgets for custom silicon to keep expanding rather than hold steady.
Its first custom AI chip has completed development and is scheduled to enter volume production in the fourth quarter of this year. A second-generation design remains on track for volume production in 2028. MediaTek expects the data-center AI chip business to generate more than $2 billion in revenue in 2026, a total that would still leave the unit well short of the phone business for now. The two-chip roadmap also gives cloud customers a second-generation path to commit to, which matters in a segment where designs are planned years in advance.
The MediaTek AI Data-Center Chip Expansion, by the Numbers
The $80 billion figure covers only custom, co-designed accelerators, not the merchant GPU market where Nvidia sells standard silicon to any buyer. MediaTek's share target therefore applies to the slice of the AI accelerator market that hyperscalers reserve for purpose-built designs, a segment defined by close co-development between chip designer and cloud customer.
| Metric | Previous guidance | Updated |
|---|---|---|
| 2027 custom AI chip addressable market | $70B to $80B | $80B |
| Target share of that market | 10% to 15% | 15% to 20% |
| First custom AI chip | In development | Volume production in Q4 2026 |
| Second custom AI chip | Planned | Volume production in 2028 |
| 2026 data-center AI chip revenue | Not disclosed | More than $2 billion |
The updated guidance shows how fast MediaTek has escalated its expectations. The 2027 market estimate now sits at the top of the previous range, and the share target implies data-center revenue growing six- to eight-fold between 2026 and 2027 if the market develops as forecast. That scale of ambition helps explain why the board set up billions in financing ahead of the production ramp.
MediaTek has not detailed how the $5 billion would be raised or what share would go to data-center programs specifically. The discretionary label means the figure is an authorization ceiling, not a spending plan. What the announcement does establish is direction: the AI data-center chip expansion tells investors and cloud customers that AI infrastructure, rather than handsets, is where the company expects its next wave of growth.
Why Mobile Is Sliding
The AI push arrives as MediaTek's traditional business weakens. Mobile-chip revenue fell 20% in the second quarter from a year earlier and 14% sequentially, with the company attributing the decline to higher bill-of-materials costs and softer smartphone demand. Global smartphone shipments dropped 11% in the quarter to their lowest level for that period since 2013, according to preliminary estimates from Counterpoint Research. The shipment figure puts the downturn in a longer frame: the industry has not seen a second quarter this weak in more than a decade, and MediaTek's response is to diversify rather than wait for a handset recovery.
The 14% sequential drop matters as much as the annual figure: it shows the weakness is ongoing rather than a one-off comparison against a strong prior-year quarter. Total second-quarter revenue still topped the company's guidance, but margins compressed as component costs climbed. Rising component costs pressure the phone business from two sides at once, squeezing both the prices MediaTek can charge and the margin it keeps on each chip.
For a company whose results have historically tracked handset cycles, the data-center pivot addresses a demand problem at the structural level. One detail in the timeline stands out: MediaTek expects more than $2 billion in data-center AI revenue in 2026 even though its first custom accelerator only enters volume production in the final quarter of the year. The projection therefore depends on a steep production ramp in the fourth quarter and on customers taking delivery quickly, which makes the coming quarter a live test of both the manufacturing schedule and real demand for the first design.
The Competitive Field: Broadcom and Nvidia
MediaTek's ASIC program puts it in direct competition with Broadcom, the established leader in custom chip design for hyperscalers, and with Nvidia's merchant data-center GPUs. Cloud providers choosing the custom route trade a multi-year development cycle for a chip tailored to their own workloads, typically at lower cost per unit once volumes climb. For cloud operators, the economics of custom silicon improve with scale: the design cost spreads across millions of units, and the chip can drop memory, interconnect, or software features the workload does not need. Custom silicon also carries a narrower software ecosystem than merchant GPUs, which is why hyperscalers pursue it mainly for their highest-volume inference workloads.
ASIC programs require heavy up-front engineering investment, with revenue arriving only after production ramps. That timing gap is why the discretionary financing matters: it lets MediaTek pursue multiple designs, including the 2028 second-generation chip, without short-term capital pressure. It also positions the company as a serious second source of custom accelerators for cloud operators that want to avoid depending on a single ASIC supplier. The $2 billion revenue expectation for 2026 is small next to the data-center GPU revenue Nvidia books each quarter, but MediaTek does not need Nvidia's scale to shift pricing dynamics in the custom segment.
For enterprise buyers and procurement teams, the practical timeline is what matters most. The first custom chip reaches volume production in the fourth quarter of 2026, and the second-generation design does not arrive until 2028, which means any real supply alternative to existing ASIC vendors is at least a year away. The strategic value of the announcement, in that sense, arrives earlier than the product: it gives cloud operators a visible second supplier to plan around.
The move also lands at a moment when AI infrastructure is the fastest-growing demand pool in semiconductors while consumer devices soften. The contrast between the two businesses, one contracting and one being built up with billions in board-level backing, is the clearest summary of where MediaTek expects the industry's growth to come from.
Why This Matters
The MediaTek AI data-center chip expansion gives cloud providers an alternative path beyond Nvidia GPUs and Broadcom designs, and it gives investors a growth story to set against a shrinking smartphone market. The raised targets set a concrete benchmark: MediaTek is now betting on $12 billion to $16 billion of custom-chip revenue by 2027, and the fourth-quarter production ramp of its first ASIC will be the first public test of that forecast.
Related Articles
- OpenAI and MediaTek Partner to Develop AI-Native Smartphone for 2027
- ByteDance AI Infrastructure Budget Hits $30 Billion for 2026
- Meta Increases 2026 Capital Expenditure to $145 Billion for AI Infrastructure
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.