MediaTek AI data centre chips

MediaTek is preparing to spend heavily on a business that looks very different from the smartphone market where it built its name.

The Taiwanese chip designer has approved a discretionary financing framework worth $5 billion. The money will support long-term growth, with custom artificial intelligence chips for data centres sitting near the centre of the plan.

This is not a straightforward promise to spend the full amount immediately. Instead, the financing gives MediaTek room to raise and deploy capital when it needs to secure production capacity, develop new hardware and compete for large cloud computing contracts.

The timing is deliberate. Smartphone demand is weakening, component costs are climbing and the market for specialised AI processors is moving in the opposite direction.

MediaTek Wants a Larger Share of the Custom AI Chip Market

MediaTek is pushing deeper into application-specific integrated circuits, better known as ASICs.

Unlike general-purpose processors, companies design these chips for particular workloads. Major cloud providers increasingly use custom silicon to run AI models more efficiently, control infrastructure costs and reduce their dependence on standard processors.

MediaTek now estimates that the addressable market for custom AI chips could reach $80 billion in 2027. It previously placed the opportunity somewhere between $70 billion and $80 billion.

The company has also raised its targeted share of that market. MediaTek is now aiming for between 15% and 20%, up from its earlier target of 10% to 15%.

That is an ambitious jump. The custom chip sector remains concentrated among a relatively small group of semiconductor companies with access to advanced manufacturing, packaging and large cloud customers.

Still, AI infrastructure spending has become too large for MediaTek to ignore.

The First MediaTek Custom AI Chip Is Nearly Ready

Chief executive Rick Tsai said MediaTek had successfully developed its first custom AI processor.

Volume production is expected to begin during the fourth quarter of 2026. A second chip remains scheduled to enter mass production in 2028.

MediaTek expects its data centre AI chip business to generate more than $2 billion in revenue during 2026. That forecast suggests the company’s move into cloud hardware is no longer a distant research project. It is already becoming a meaningful commercial operation.

The $5 billion financing framework gives MediaTek flexibility around that growth. Advanced AI chips require more than design work. The company may need to secure leading-edge manufacturing capacity, advanced packaging, high-speed interconnect technology and other parts of an increasingly crowded semiconductor supply chain.

MediaTek has already said it plans to increase investment in technologies such as silicon photonics, co-packaged optics and faster SerDes interfaces. It is also working with Taiwan Semiconductor Manufacturing Company on advanced manufacturing and packaging technologies for future data centre products.

Smartphone Weakness Is Making Diversification More Urgent

MediaTek’s traditional mobile chip business is facing a rougher market.

Revenue from mobile processors fell 20% year on year during the second quarter. Higher memory and component costs have pushed up handset prices and weakened demand, particularly in more price-sensitive parts of the smartphone market.

Preliminary Counterpoint Research estimates showed global smartphone shipments falling 11% during the quarter, reaching their lowest level for that period since 2013. MediaTek expects global smartphone unit shipments to decline by around 15% across 2026.

The company is responding partly through price increases. Tsai said MediaTek would adjust product pricing to reflect higher costs throughout the supply chain.

That may protect margins, but it does not solve the larger problem. Smartphones remain a mature and highly competitive market. Data centre AI chips offer faster growth, larger contracts and potentially stronger margins.

MediaTek does not need to abandon mobile processors. It needs another major source of revenue beside them.

Quarterly Revenue Grew, but Profit Fell

MediaTek reported second-quarter revenue of NT$152.18 billion, equivalent to about $4.71 billion. Revenue increased 1.2% compared with the same period a year earlier.

Net income moved the other way, falling 12.3% to NT$24.6 billion.

Investors appeared more interested in what comes next. MediaTek shares climbed 9.9% on 31 July before the earnings announcement. The stock had risen 148.6% since the beginning of 2026, compared with a 48.9% increase in Taiwan’s benchmark index.

That rally carries expectations. MediaTek will now have to show that its custom AI business can move from promising forecasts to reliable production and repeat orders.

MediaTek Is Trying to Become More Than a Phone-Chip Company

The broader strategy is becoming clearer.

MediaTek spent years building scale through processors used in smartphones, televisions, wireless devices and consumer electronics. AI data centres require a different level of engineering, customer support and supply-chain coordination.

They also offer a much larger prize.

Cloud companies are spending heavily on their own chips as they search for alternatives to expensive, power-hungry general-purpose AI hardware. Custom processors can be tuned for particular models, inference workloads or internal cloud services.

MediaTek believes it can claim part of that market. Its $5 billion financing framework gives the company the capacity to move quickly when manufacturing slots, packaging capacity or major customer programmes become available.

The difficult part starts now. Producing one custom chip is not the same as building a lasting data centre business. MediaTek must deliver on schedule, secure follow-up designs and prove that it can compete with semiconductor companies already deeply embedded in cloud infrastructure.

Still, this is no small side project. MediaTek is putting billions of dollars behind a bet that its next phase of growth will happen inside AI data centres—not inside another generation of smartphones.

Sources