ASUS server business grows 200% as Intel, AMD and edge AI widen its infrastructure bet

ASUSTeK Computer Inc. (Taiwan Stock Exchange: 2357) is widening its artificial-intelligence infrastructure strategy from large data-center servers into a multi-vendor ecosystem spanning cloud computing, enterprise AI and rugged industrial edge systems, after its server business grew about 200% year over year during the second quarter of 2026. At ASUS AI Tech 2026 in Seoul, the company showcased new and existing infrastructure built around Intel Xeon 6, next-generation Intel server processors, AMD EPYC 9006 CPUs and AMD Instinct MI350P accelerators, alongside edge systems capable of up to 180 AI TOPS. Technology partners including Aleria, Foxlink, IBM, Samsung, Schneider Electric and Western Digital were positioned around the broader infrastructure stack, reflecting ASUS’ attempt to compete on complete AI deployment rather than servers alone. The expansion arrives as management lifts its full-year AI server growth expectation to around 150%, well above its original 50% to 100% target. The strategic tension is that ASUS is rapidly becoming a meaningful AI infrastructure supplier, but approximately 90% of its second-quarter server revenue still came from NVIDIA HGX and GB300 NVL72 systems, making supplier diversification and higher-value integration critical to the next stage of growth.

The Seoul event is therefore less a completely new product launch than a statement about how ASUS wants its infrastructure business to evolve. Several technologies being highlighted, including the RUC-2000 industrial edge system, were introduced earlier in 2026, while ASUS has already been shipping high-end NVIDIA-based AI systems and expanding its Infrastructure Solutions Business Group. The new development is the breadth of the ecosystem being assembled around those products, from data-center CPUs and accelerators through memory, storage, power, cooling, AI governance and edge deployment. If ASUS can turn that ecosystem into repeatable enterprise and cloud deployments, it could capture more of the economics surrounding AI infrastructure rather than remaining primarily a hardware assembler riding demand for whichever accelerator platform is currently strongest.

Why is ASUS expanding its Intel and AMD server lineup when NVIDIA already drives most AI server revenue?

ASUS’ current AI server momentum is heavily linked to NVIDIA. During its second-quarter investor discussion, management said HGX-based products represented approximately 60% of server revenue and GB300 NVL72 rack-scale systems contributed around 30%, leaving only roughly 10% for general-purpose servers and other products. That means about nine dollars out of every ten dollars of current server revenue are tied to NVIDIA-centered architectures.

The dependence is not inherently negative because NVIDIA remains at the center of the global AI infrastructure investment cycle, and ASUS has benefited substantially by moving early with Blackwell and Blackwell Ultra systems. Its AI POD based on NVIDIA GB300 NVL72 combines 72 Blackwell Ultra GPUs with 36 Grace CPUs, while the company has also prepared infrastructure based on the next-generation NVIDIA Vera Rubin platform. ASUS has additionally invested in liquid cooling, networking, deployment software and professional services needed to install increasingly complex rack-scale systems.

The Seoul portfolio nevertheless shows why management does not want its infrastructure opportunity defined by one accelerator supplier. ASUS is highlighting RS700-E12-RS4U and RS720-E12-RS12U systems built around Intel Xeon 6 processors, alongside the XA P8I-E13A high-density 6U platform designed for Intel’s next-generation server processors and accelerated AI workloads. These products give customers additional CPU-based choices for enterprise computing, high-performance computing and data-intensive applications where a full NVIDIA rack may be unnecessary or uneconomic.

AMD adds another route. The RS720A-E14B-R32U and RS500A-E14B-R12U are built around AMD EPYC 9006 processors, while the ESC8000A-E13P can use AMD Instinct MI350P PCIe accelerators for inference, agentic AI and high-performance computing. ASUS is therefore positioning itself as an infrastructure integrator capable of deploying whichever compute architecture best matches workload, power, budget and software requirements rather than requiring customers to standardize every AI project around one vendor.

Can ASUS turn a 200% server revenue surge into a durable business rather than a single AI hardware cycle?

The growth rate gives ASUS considerable room to make that case. Management says the server business expanded around 200% year over year and approximately 58% sequentially during the second quarter, prompting the company to increase its full-year 2026 AI server growth expectation to around 150%. The original goal at the beginning of the year had been growth of 50% to 100%, so the revised outlook implies demand has materially exceeded management’s initial assumptions.

The wider financial results support the scale of the acceleration. ASUSTeK Computer reported consolidated first-half revenue of NT$466.1 billion, up approximately 39% from NT$335.7 billion a year earlier. Second-quarter revenue can be derived at about NT$257.7 billion, roughly 37% above the NT$188.0 billion reported in the corresponding 2025 quarter.

Profit growth was even stronger during the second quarter. Consolidated net income for the first six months reached NT$31.5 billion, up about 31% year over year. After subtracting first-quarter earnings, second-quarter net income was approximately NT$20.3 billion, almost 95% above the comparable NT$10.45 billion generated a year earlier.

Those numbers show ASUS is benefiting from more than a conceptual AI narrative. The challenge is sustaining growth as accelerator generations change rapidly and customers continuously reassess architecture. A server manufacturer that wins primarily because it has access to the latest GPU can lose momentum when supply normalizes or another supplier becomes more competitive, which explains ASUS’ growing emphasis on engineering, deployment, software and lifecycle services around the underlying silicon.

Why is ASUS building an AI Factory stack instead of competing only on server hardware specifications?

Modern AI infrastructure is increasingly difficult to sell as a collection of independent servers. High-density accelerators create demanding requirements for networking, storage, power delivery, cooling, rack design, software orchestration and system monitoring. A customer building hundreds or thousands of accelerators into an AI factory needs those components to operate as one environment, making integration expertise increasingly valuable.

ASUS has been moving toward that model through its AI Factory architecture and Infrastructure Solutions Business Group. The company combines compute hardware with high-speed storage, networking, liquid cooling, ASUS Infrastructure Deployment Center software and ASUS Control Center management tools. It also offers professional services intended to support design, validation, deployment and ongoing operation of multi-node and multi-rack AI clusters.

That broader stack potentially improves two commercial variables. First, it can shorten customer deployment time because hardware, cooling, networking and software are validated together rather than assembled independently after delivery. Second, it gives ASUS more opportunities to create value beyond the server itself, an important consideration because competition among original equipment manufacturers can pressure hardware margins even during periods of exceptionally strong demand.

Partners highlighted at ASUS AI Tech 2026 reinforce this approach. Samsung and Western Digital represent important parts of the memory and data-storage ecosystem, while Schneider Electric operates across power management and cooling infrastructure. IBM contributes enterprise technology and data-management capabilities, while other partners extend the stack into software, components and governance. ASUS does not need to manufacture every layer if it can function as the company responsible for integrating them into deployable infrastructure.

Why could industrial edge AI become the second growth engine beside massive cloud AI factories?

ASUS is simultaneously pushing compute in the opposite direction. While its rack-scale systems concentrate enormous processing power inside data centers, the RUC-2000 series is designed to place AI inference directly inside factories, vehicles, healthcare environments, surveillance installations and other locations where sending every data stream to the cloud may be impractical.

The RUC-2000H uses Intel Core Ultra Series 3 processors and delivers up to 180 AI TOPS in a fanless half-rack industrial design. It supports one 10-gigabit Ethernet connection, five 2.5-gigabit Ethernet ports, up to eight GMSL2 camera interfaces and expansion for a discrete GPU of up to 200 watts. ASUS says the system can operate between minus 25 degrees Celsius and 70 degrees Celsius and has been validated against MIL-STD-810H requirements.

Those specifications address a different problem from cloud AI training. A manufacturing line inspecting components through multiple cameras cannot always tolerate latency from continuously transmitting raw video to a distant data center. Hospitals may prefer to keep sensitive patient information inside the facility, while vehicles, transportation systems and public-safety infrastructure can require local decisions even if external connectivity becomes unreliable.

Edge inference therefore creates demand for hardware that combines computing power with environmental resilience, specialized connectivity and long product availability. Standard rack servers built for temperature-controlled data centers do not automatically satisfy those requirements. ASUS’ background in motherboards, industrial computing, PCs and servers gives it several engineering capabilities that can be combined around this market.

The strategic attraction is also diversification. Training enormous models requires a relatively concentrated group of cloud providers, sovereign AI projects and large enterprises able to finance expensive accelerator clusters. Edge AI potentially distributes inference across much larger numbers of factories, medical sites, vehicles and industrial systems, creating a different volume opportunity as trained models move into real-world applications.

Does ASUS have a credible cloud-to-edge advantage or is the portfolio becoming too broad?

The range is both the strength and risk of ASUS’ strategy. The company now describes its AI opportunity across infrastructure, physical AI, AI PCs and devices, and software-enabled applications. At the infrastructure level alone it spans giant liquid-cooled racks, conventional enterprise servers, storage, workstations, deskside AI supercomputers and rugged industrial computers.

Breadth can reduce dependence on a single product cycle because the same AI investment wave appears in different forms. A model may be trained in a GPU cluster, fine-tuned on a smaller enterprise system and eventually deployed on workstations or industrial devices at the edge. ASUS wants to supply hardware and management layers at several points along that workflow.

The danger is that breadth does not automatically create integration. Enterprises will not choose an ASUS data-center server simply because the company also manufactures an edge computer, and an industrial customer may not care which brand of rack trained the model being deployed in its factory. The commercial advantage appears only when ASUS can connect those environments through common deployment, management, security or application capabilities.

Management is explicitly trying to build that connective layer. Co-chief executive S.Y. Hsu recently described ASUS’ strategy as combining AI devices, infrastructure and applications from cloud to edge rather than defending traditional hardware against software-driven change. ASUS Zenni Claw, ASUS AI Hub and other software initiatives are intended to extend that strategy beyond physical products.

The next evidence investors need is therefore less about how many product categories ASUS can list and more about whether enterprises increasingly buy integrated packages containing infrastructure, software and services. Customer deployments, recurring software revenue and professional-services contribution would provide stronger evidence of full-stack value capture than another expansion of the hardware catalogue.

How does ASUS compare with Dell, Supermicro and other companies chasing AI infrastructure spending?

ASUS is pursuing one of the most competitive areas in global technology. Dell Technologies has been reporting extraordinary AI server demand and recently lifted its fiscal-year AI-optimized server revenue outlook to approximately $74 billion after receiving more than $130 billion of AI server orders over the preceding year. Foxconn is also expanding production around NVIDIA’s next-generation platforms while Super Micro Computer and multiple Asian original design manufacturers compete aggressively for hyperscale and enterprise deployments.

ASUS does not need to become the largest supplier for the infrastructure strategy to materially change its own financial profile. Its server business was historically much smaller than its PC and component franchises, so several years of triple-digit or high double-digit growth can shift the group’s revenue mix even if ASUS remains smaller than Dell or the largest contract manufacturers.

Its potential advantage lies in combining a recognized commercial technology brand with engineering across motherboards, PCs, servers and industrial systems. ASUS can also support Intel, AMD and NVIDIA architectures rather than making a single proprietary processor bet. The company’s work on national-scale supercomputing projects and sovereign AI infrastructure gives it references beyond traditional enterprise IT.

Scale remains important, however. Large AI customers prioritize supply assurance, rapid access to new accelerators, global service capacity and the ability to deploy thousands of nodes without interruption. As ASUS pursues larger rack-scale orders, supply-chain execution and working-capital requirements can become as important as technical design.

What does the ASUS share-price rally say about expectations already built into AI server growth?

ASUSTeK Computer shares closed September 3 at NT$971, down approximately 3.9% for the session. That decline should not be attributed to the Seoul announcement because the international release was distributed after Taiwan’s regular trading session had already ended.

The stock was effectively flat compared with its August 27 close of NT$968, but had gained approximately 21.5% from NT$799 on August 3. Its 52-week range stood at roughly NT$490 to NT$1,025, leaving the September 3 close only about 5% below the high and nearly double the lower end of the range.

The recent rerating followed strong second-quarter results and management’s higher AI server growth expectations, meaning investors have already assigned considerable value to the infrastructure acceleration. That raises the bar for future announcements because additional product launches may have less valuation impact than evidence that growth remains strong while profitability and market share hold up.

The next quarterly results will therefore matter more than the September showcase in determining whether the rerating is sustainable. Server growth above 150%, further expansion of rack-level shipments and evidence that new Intel, AMD and edge platforms are producing actual revenue would support the thesis that ASUS is broadening successfully beyond its current NVIDIA-heavy mix. Slower order conversion, component shortages or declining profitability as server competition intensifies would weaken it.

What will prove that ASUS has built a real AI infrastructure ecosystem from cloud to edge?

The first proof point is supplier diversification. With NVIDIA HGX and GB300 systems contributing around 90% of second-quarter server revenue, meaningful sales from AMD Instinct, Intel-based infrastructure and general-purpose enterprise platforms would show that the wider portfolio is producing revenue rather than merely providing customers with theoretical choice.

The second is continued server growth after the extraordinary 2026 comparison base becomes harder. Management’s expectation of around 150% AI server growth for the full year already represents a major acceleration. Maintaining strong growth during 2027 would show that ASUS is capturing an enduring share of enterprise infrastructure spending rather than benefiting primarily from one Blackwell deployment wave.

The third is movement beyond hardware revenue. Adoption of ASUS AI Factory deployment software, professional services, management tools and integrated partner solutions would indicate that the company is capturing more value per project and making itself harder to replace with another server vendor.

Industrial deployments provide the fourth test. The RUC-2000 series gives ASUS a credible edge-computing platform, but meaningful customer programs in manufacturing, healthcare, transportation or public safety would demonstrate whether the cloud-to-edge thesis exists outside trade-show demonstrations.

ASUS enters that test from a much stronger position than it held a year ago. Server revenue has roughly tripled year over year, management has lifted its growth target sharply, first-half consolidated revenue has risen almost 39% and the stock is trading near the upper end of its 52-week range. ASUS AI Tech 2026 shows how management intends to build on that momentum: reduce architectural dependence through Intel and AMD, surround servers with storage, cooling and software partners, and extend inference into industrial environments. The next strategic proof will come when those additional layers begin contributing enough revenue to show that ASUS is becoming an AI infrastructure integrator rather than simply one of the major beneficiaries of NVIDIA’s current server cycle.

What are the key takeaways from ASUS’ cloud-to-edge AI infrastructure expansion?

• ASUSTeK Computer showcased an expanded AI infrastructure ecosystem at ASUS AI Tech 2026 in Seoul.
• The portfolio includes server platforms based on Intel Xeon 6, next-generation Intel processors, AMD EPYC 9006 CPUs and AMD Instinct MI350P accelerators.
• Partners highlighted around the ecosystem include Aleria, Foxlink, IBM, Samsung, Schneider Electric and Western Digital.
• ASUS’ RUC-2000 industrial edge platform delivers up to 180 AI TOPS using Intel Core Ultra Series 3 processors.
• ASUS says its server business grew about 200% year over year and approximately 58% sequentially during the second quarter.
• Management has lifted its full-year 2026 AI server growth expectation to around 150%, compared with its original 50% to 100% target.
• NVIDIA HGX systems represented about 60% of second-quarter server revenue and GB300 NVL72 systems around 30%, leaving ASUS’ current server business highly exposed to NVIDIA architectures.
• ASUSTeK Computer’s first-half consolidated revenue increased approximately 39% to NT$466.1 billion, while net income rose about 31% to NT$31.5 billion.
• ASUS shares closed September 3 at NT$971, around 21.5% above their August 3 close and roughly 5% below the 52-week high.
• The next evidence of successful diversification will be meaningful revenue from Intel, AMD, edge AI, software and services alongside continued growth in ASUS’ NVIDIA-based systems.