Blog

DEEP RESEARCH · GAONCHIPS/AI ASIC

[Semiconductors] Design-Solution Company: Gaonchips

A semiconductor research note on AI ASIC demand and Samsung Foundry design-solution partners

Date: 2025-02-08 · Semiconductor/design-house view · Naver Blog

Investment decisions are your own responsibility. This material is research and is not a buy or sell recommendation.

0. Bottom line first

AI ASIC demand is structurally expanding across data centers, edge AI, and autonomous driving. From a Korean equity perspective, Samsung Foundry's advanced process roadmap, SAFE ecosystem, and DSPs such as Gaonchips and SemiFive are the channels that could absorb this demand.

AI ASIC demandLower GPU dependence and better power efficiency
Data centersTPU, Trainium, MTU and custom accelerators
Edge/IoTMobile NPUs, smart devices, industrial IoT
AutomotiveADAS, FSD, automotive AI SoCs
DSPsSamsung SAFE-based design support
Custom AI-chip demand flows into foundries and design-solution companies

1. Growth drivers for ASICs

Official fact: The post cites a Morgan Stanley view that the AI ASIC market could grow from roughly $12 billion in 2024 to $30 billion in 2027. It also notes a 38% annual growth rate for the overall AI-chip market over the same period.

Interpretation: The core trend is that hyperscalers are not relying only on NVIDIA GPUs, but are adding custom ASICs such as Google TPU, Amazon Inferentia/Trainium, and Meta MTU. ASICs designed for specific workloads can have advantages in power efficiency and cost.

Official fact: The post says the edge AI chip market is expected to grow from about $16 billion in 2024 at more than 30% annually to $150 billion by 2030. It also presents the automotive AI market as growing at a 20%-plus annual pace into the mid-to-late 2020s.

  • Data centers: custom accelerators are needed for generative-AI training and inference.
  • Edge/IoT: smartphones, smart appliances, and industrial IoT devices need low-power specialized AI chips.
  • Autonomous driving: Tesla FSD and ADAS require large amounts of compute, creating demand for automotive ASICs.

2. Samsung Foundry and the DSP ecosystem

Official fact: At SFF 2024, Samsung presented AI/HPC-focused GAA processes, HBM stacking, high-speed packaging, SF2Z and SF4U, and an AI Solutions platform that combines design, memory, and packaging.

Interpretation: AI ASIC customers need to solve design complexity, verification, memory, and packaging together. That makes not only Samsung Foundry's process capability, but also the quality of the SAFE design ecosystem and DSP partners, important.

SemiFive

Platform-based SoC design

A Korean fabless company founded around 2019 by former SiFive people and a Samsung SAFE DSP partner. The source notes three AI-inference SoC platforms, eight custom ASIC tape-outs, 14nm AI inference SoC mass production, a Samsung 5nm HPC NPU plan, and a HyperAccel 4nm LPU collaboration.

Gaonchips

ARM and Samsung turnkey design

A system semiconductor design company founded in 2012, described as a Samsung Foundry Best Design Partner and an ARM Approved Design Partner. Its strengths include ARM CPU/GPU/NPU IP, 14nm-to-5nm tape-out experience, and Squire bitcoin-mining ASIC mass-production test experience.

3. Key player comparison

CompanyRoleAI ASIC pointPartners/customers
SemiFivePlatform-based ASIC design14nm AI inference chip, 5nm HPC NPU, HyperAccel LPU collaborationSamsung Foundry, Arm, HyperAccel
GaonchipsTurnkey design serviceARM IP-based SoCs, 5-to-14nm design experience, mining ASIC track recordSamsung Foundry, Arm, North American and Japanese customers
BroadcomGlobal custom ASIC leaderHyperscaler custom AI ASICs including Google TPU; the post cites FY2023 AI-related revenue of $12.2 billion, up 220%Google, Meta, OpenAI and others
Qualcomm/Tesla etc.Edge and automotive AI chipsCloud AI 100, Snapdragon AI Engine, Tesla FSDData-center, mobile, and automotive ecosystems

4. Gaonchips investment angle

Interpretation: For Korean investors, the key point is Gaonchips' position as a Samsung and ARM partner. It can be a channel for domestic fabless AI ASIC demand, and its strategic value could rise if Samsung Foundry strengthens partner support to expand its customer base.

The post also assumes that Gaonchips' revenue base is still small. The opportunity comes from possible project wins, but actual reference cases and customer acquisition still need confirmation.

5. Risk factors

  • Semiconductor cycle: If generative-AI demand cools or data-center capex is delayed, ASIC orders can fall.
  • Technology and ecosystem gap: ASICs are optimized for specific workloads but less flexible, and they must compete with software ecosystems such as NVIDIA CUDA.
  • Manufacturing dependence: TSMC or Samsung process delays, yield issues, and advanced-foundry supply-chain risks transfer directly to design houses and customers.
  • Customer concentration: As with Broadcom and Google TPU, large customers internalizing design or switching suppliers can raise volatility. SemiFive and Gaonchips also depend heavily on Samsung Foundry.
  • Intensifying competition: AMD, Intel, big-tech internal teams, and startups are all entering the market, creating possible oversupply and margin pressure.

AI ASICs are a large opportunity, but only companies that can manage demand uncertainty, technology shifts, supply chains, customer relationships, and competition will capture the upside. The current candidates are global Broadcom and, domestically, Samsung Foundry and its partners.

Sources