Tuesday, March 4, 2025

Nvidia Wants to Move in Adjacencies; So Do Many in the AI Value Chain

Participants in the artificial intelligence value chain can be expected to eventually begin encroaching on other roles. Much of that already is happening. Large end users already design their own acceleration chips.


And Nvidia’s business ambitions already extend far beyond being a supplier of graphical processing units and seem primarily focused on becoming a supplier of Artificial Intelligence as a Service, high-performance computing and possibly some parts of the enterprise AI solutions market.


Nvidia is not trying to be a general-purpose cloud provider like AWS or Azure and  does not want to compete in general cloud hosting (compute, storage, databases) like AWS EC2, Azure Blob Storage, or Google Cloud Functions.


Nor does it want to become a supplier of enterprise solutions such as Microsoft’s productivity tools. 


On the other hand, HPC and AIaaS offer lots of room for expansion within the ecosystem.  


Artificial Intelligence as a Service (AIaaS) Market

udy

Estimated Market Size (2024)

Projected Market Size

CAGR

Projection Year

Source

Allied Market Research

$11.7 billion

$178.9 billion

35.90%

2032

alliedmarketresearch.com

Grand View Research

$16.08 billion

Not specified

36.10%

2030

grandviewresearch.com

MarketsandMarkets

Approximately $14.00 billion

$72.13 billion

38.80%

2029

marketsandmarkets.com

Precedence Research

$11.96 billion

$294.83 billion

37.78%

2034

precedenceresearch.com



High-Performance Computing (HPC) Market

Study

Estimated Market Size (2024)

Projected Market Size

CAGR

Projection Year

Source

Fortune Business Insights

$54.39 billion

$109.99 billion

9.20%

2032

fortunebusinessinsights.com

MarketsandMarkets

Not specified

$49.9 billion

6.70%

2027

marketsandmarkets.com

BCC Research

Not specified

$107.8 billion

15.60%

2028

blog.bccresearch.com


“AI as a service” is probably the most pointed challenge to Nvidia’s “cloud computing as a service” customers. DGX Cloud directly competes with the cloud providers’ AI “as a service” offerings.


Nvidia provides a “full-stack” AI solution based on frameworks (CUDA, TensorRT, NeMo) optimized for AI workloads.


Production of custom chips that could compete with AWS Trainium and Inferentia;

Google Tensor Processing Unit or Microsoft Maia AI chips is another potential area of competition. 


Beyond that, Nvidia is selling turnkey AI supercomputers (DGX SuperPODs), which cloud providers can use to build AI infrastructure. In some cases, such clusters might be sold directly to enterprises.


Nvidia’s GeForce NOW is a niche business that competes with Microsoft’s xCloud and AWS Gamelift in the cloud gaming space.


Over time, many contestants in one part of a computing value chain have moved into adjacent areas. 


Company

Original Role

Expansion Move

Adjacent Market Entered

Product/Service

Year

Intel

Semiconductor (CPU) Maker

Launched discrete GPUs

Graphics Processing (Competing with Nvidia & AMD)

Intel Arc GPUs

2022

Nvidia

GPU Maker

Developed AI Cloud Services

AI as a Service (Competing with AWS, Google, Azure)

Nvidia DGX Cloud

2023

AMD

CPU & GPU Maker

Acquired Xilinx

FPGA & Custom Silicon (Data Centers, AI, Automotive)

Xilinx Adaptive SoCs & AI Engines

2022

Amazon (AWS)

Cloud Services

Designed Custom AI Chips (Inferentia & Trainium)

AI/ML Hardware (Competing with Nvidia)

AWS Trainium & Inferentia Chips

2018+

Google (GCP)

Cloud Services & Search

Developed Custom AI Chips (TPUs)

AI/ML Hardware (Competing with Nvidia)

Google Tensor Processing Unit (TPU)

2016+

Microsoft (Azure)

Cloud Services

Launched Custom AI & ARM Chips

AI & ARM Computing (Competing with AWS, Nvidia)

Azure Maia AI Chips & Cobalt CPUs

2023

Apple

Consumer Electronics

Designed Custom ARM Chips

Semiconductor (Replacing Intel CPUs)

Apple M1, M2, M3 Chips

2020+

Tesla

EV Manufacturer

Developed AI Supercomputers (Dojo)

AI & Computing (Competing with Nvidia for AI training)

Tesla Dojo AI Training Cluster

2023+

Oracle

Database Software

Expanded Cloud & AI Infrastructure

Cloud Computing (Competing with AWS, Azure)

Oracle Cloud Infrastructure (OCI)

2016+

Meta (Facebook)

Social Media

Designed Custom AI & ML Chips

AI Hardware & Compute (Competing with Nvidia)

Meta Training & Inference Accelerator (MTIA)

2023

No comments:

How Much Will AI Compute Grow to 2030, Compared to Electricity Consumption?

A report by the International Energy Administration estimates data centers accounted for around 1.5 percent of the world’s electricity cons...