AI

AI市場の最新動向や注目トレンドを、包括的な業界分析を通じてお届けします。市場環境の変化、新たなテクノロジー、競争動向を継続的に捉えながら、半導体からクラウド、アプリケーションまで、AIエコシステム全体を幅広くカバーしています。さらに、AI市場の成長を支える要因や、今後のマネタイズの可能性についても分析。データに基づく調査と実践的なインサイトを通じて、より確かな企業の意思決定を支援しています。

"AI" Data and Research

Every AI dollar starts as a wafer and ends as a token.

We track the whole path. Foundry wafers and CoWoS capacity, compute silicon, HBM, optics, servers, racks, infrastructure spends, AI token demand, economics to use-cases — modelled in one connected dataset, refreshed quarterly, with analyst deep dives on every layer.

1. Foundry & Packaging
Where capacity, not demand, sets the ceiling. Wafer consumption is built bottom-up from every AI compute vendor's roadmap, then checked against node and packaging capacity.
- Wafer shipments by chip vendor and process node
- CoWoS-S/R/L, EMIB-T, etc. allocation by vendor
- Yields, utilisation rates and foundry–vendor roadmaps
- Foundry allocation across TSMC, Samsung and Intel Foundry
Covered in-depth further · Foundry Service

1. Foundry & Packaging

Where capacity, not demand, sets the ceiling. Wafer consumption is built bottom-up from every AI compute vendor's roadmap, then checked against node and packaging capacity.

Wafer shipments by chip vendor and process node
CoWoS-S/R/L, EMIB-T, etc. allocation by vendor
Yields, utilisation rates and foundry–vendor roadmaps
Foundry allocation across TSMC, Samsung and Intel Foundry

Covered in-depth further · Foundry Service

Every layer is delivered twice, on a quarterly or semi-annual cadence: once as a sliceable quantitative tracker and forecast, once as in-depth qualitative analysis. Subscribers get both, so a shipment revision always comes with the reason behind it.

Questions this answers

Supply & Capacity

Is there enough wafer and packaging capacity, and are utilisation trends strong enough, to support announced AI compute roadmaps — and who gets squeezed first if not?
How is wafer capacity and advanced packaging allocated between vendors, and how does it shift each quarter?
What are the wafer pricing trends and the future foundry process node and technology roadmaps?
What are the memory allocation, pricing and capacity trends by technology – LPDDR vs HBM – by supplier?

Compute & Competition

How is demand splitting between merchant GPUs and hyperscaler custom ASIC, and how fast is the mix moving?
Which custom silicon programmes are reaching volume, and which are quietly slipping?

Content per Rack

What is the memory, optics, power and cooling content per accelerator, server and rack — and the trend?
How fast is liquid cooling displacing air, and who captures the dollar content? What are the key trends?
What does an AI rack cost, layer by layer, and where is it most exposed to component pricing?

Power & Interconnects

How much compute can be delivered inside a fixed power envelope, and what is that worth?
How do rack-scale capabilities and deployments compare between NVIDIA, AMD, Google & others?
How will scale-up, scale-out and scale-across evolve from copper to optical interconnect technologies?

Demand & Monetisation

How is CSP capex allocated between in-house and commercial silicon?
Do current token economics support the capex being committed, and at what utilisation does the case break?
Is token demand outpacing token generation?
How are tokenomics changing between inference and training?
Which are the top AI Cities? Is Sovereign AI taking off? Who is winning and who is ahead?

Latest Research

Figure AI: Assessing the Race for General-Purpose Humanoids

Overview: Figure AI is emerging as a leading player in humanoid robotics, combining proprietary AI, purpose-built hardware and in-house manufacturing to build general-purpose robots. This report examines its technology stack, commercial progress and competitive positioning, while assessing whether its $39 billion valuation can be justified by real-world adoption.It takes a closer look at Helix, Figure 03 and the BotQ manufacturing facility to understand how AI capabilities, hardware design and production scale could translate into a lasting competitive advantage. The report also explores the challenges ahead, from expanding commercial deployments to competing with Chinese players and Tesla. Ultimately, it assesses whether Figure can turn its technological promise into a scalable business over the next two years.

2026年10月22日

Insight

PRs & Blogs

Wearables