Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →No specific under-the-radar stock can be identified from the available evidence behind this headline. AMD and Broadcom are discussed as alternatives to Nvidia, but neither is established as the unnamed company—and neither can be called a likely winner on that basis. What can be assessed is how their different businesses fit into the AI infrastructure race and what investors would need to verify before backing any supposed dark horse.
Why the headline does not identify a stock
“Forget Nvidia and AMD” signals that a third company is the subject, but the available material does not name that issuer or establish a company-specific investment case. Broadcom appears in related commentary as a potential alternative to Nvidia and AMD; that is not evidence that Broadcom is the stock the headline means. Naming it—or another company—would turn an unresolved identity into an unsupported claim.
As an Amazon Associate I earn from qualifying purchases.
Nor does the evidence show that any one company will “win” the AI semiconductor race. That outcome depends on distinct products, workloads, customers, software, execution, and financial results. The useful question is not simply which chipmaker has the best AI story, but which company can serve a particular workload competitively and convert that position into durable business results.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsAI chips compete across different jobs
“AI chip” is an umbrella term, not one interchangeable product category. Training large models, running inference, coordinating data-center workloads, and moving data through a network call for different components. A company can benefit from AI infrastructure demand without making the general-purpose GPUs most associated with AI training.
#1 Best Overall
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
- Training GPUs perform the intensive computation used to train models. They are also used for some inference workloads.
- Inference GPUs run models after training. Cost and efficiency can matter especially strongly when deployments serve large volumes of queries.
- Data-center CPUs handle general computing and can coordinate workloads and data-management tasks alongside accelerators.
- Custom ASICs are designed for specific tasks or customers. They can trade the flexibility of a general-purpose GPU for performance or efficiency on selected workloads.
- Networking and interconnects move data among chips and systems. They can be important to scaling an AI deployment even though they are not themselves model-training processors.
How the named companies differ
| Company | Exposure described in the commentary | What the investment thesis emphasizes | Important qualification |
|---|---|---|---|
| Nvidia | AI infrastructure and GPUs, with the CUDA software ecosystem. | Its combination of hardware and software is presented as a competitive advantage. | The commentary’s characterization does not establish current market share or guarantee that Nvidia will retain its position. |
| AMD | GPUs and data-center CPUs. | The cited thesis emphasizes possible growth in inference, where cost and efficiency matter, and CPU demand if AI agents increase the need for workflow coordination and data management. | These are analyst arguments about potential opportunities, not verified predictions of demand or market outcomes. |
| Broadcom | Custom ASIC design and data-center networking. | The thesis focuses on customer-specific chips for selected workloads and the infrastructure connecting data-center systems. Commentary also discusses Broadcom’s role in Google/Alphabet TPU development. | ASICs are not interchangeable with general-purpose GPUs for every workload. The cited customer example and growth opportunity do not by themselves establish future revenue or market leadership. |
The comparison is about business exposure, not a ranking. Nvidia’s software ecosystem is part of the competitive picture; alternatives also have to meet the performance, cost, efficiency, flexibility, and software needs of the particular deployment. A custom chip that fits one customer’s workload is not automatically a general substitute for GPUs across the market.
What the cited revenue figures do—and do not—show
Several large Broadcom figures appear in the related commentary, but they describe estimates or forecasts rather than a confirmed result for an unnamed stock:
Rank #2
- Delivers 25 TFLOPS of AI inference performance using four MemryX MX3 accelerators in a compact M.2 2280 M-key module.
- Connects through PCI Express Gen 3 and above for straightforward integration into embedded and edge systems with an M.2 M-key slot.
- Includes 42 MB of distributed on-chip weight storage to support low-latency, batch-1 pipelined inference.
- Designed for low-power deployment with a 1 W–12 W operating range and host-managed airflow requirements.
- Supports ONNX, TensorFlow, Keras, and TFLite models through the MemryX software stack on Linux and Windows hosts.
- A 2026 Motley Fool article reported Broadcom’s projection that AI ASIC revenue would exceed $100 billion in fiscal 2027. This is a forward-looking company projection as reported by a secondary source, not realized revenue.
- A 2025 Motley Fool article described a potential total addressable opportunity of up to $90 billion from three advanced AI chip customers by 2027. That is an opportunity estimate, not a revenue result.
- A January 2026 Motley Fool article reported just under $64 billion in Broadcom total revenue in the prior year, with about $20 billion related to AI. It also attributed forecasts of more than $50 billion in AI revenue in the current fiscal year and more than $100 billion in fiscal 2027 to Citi analysts—not to Broadcom guidance. The fiscal-period definitions and underlying figures are not established here.
These numbers have different sources, periods, and meanings. A projection, a potential addressable market, and an analyst forecast cannot be treated as interchangeable measures of sales. None identifies the unnamed company in the headline, and none demonstrates that Broadcom—or any competitor—will win.
Partnerships and standards are evidence to track, not a verdict
Related commentary describes AMD GPU deployments planned with Oracle and OpenAI in a January 2026 article. Those are reported plans, not proof of completed deployments, sustained demand, or a particular financial outcome. The same commentary describes Broadcom’s role in Google/Alphabet TPU development; that customer example does not establish how much revenue it generates or how broadly a custom-chip model will spread.
AMD is also described as participating with Broadcom and Intel in the UALink Consortium, an effort to develop an open interconnect standard. Its success is prospective and long-term: the material does not show that UALink has replaced Nvidia’s NVLink. Participation in a standards effort is a development to monitor, not evidence that a new standard will gain adoption.
What an investor should check before accepting a “dark horse” thesis
Once the company is actually identified, judge the claim against company filings and other primary disclosures rather than relying on a headline or an analyst’s opportunity estimate. A useful diligence checklist is:
Rank #4
- ✅Powered by 26 Tera-Operations Per Second (TOPS) Hailo-8 AI Processor. 2.5W typical power consumption
- ✅Scalable, enabling simultaneous processing of multi-streams & multi-models
- ✅Enabling real-time, low latency and high-efficiency AI inferencing on the edge devices
- ✅Supports TensorFlow, TensorFlow Lite, ONNX, Keras, Pytorch frameworks
- ✅Supports Linux and Windows. Supports the temperature range of -40°C to 85°C
- Workload and product: Does the company sell training or inference GPUs, CPUs, custom ASICs, networking equipment, or several of these? Which workload is driving the claim?
- Customer evidence: Are deployments completed, planned, or merely possible? Is demand tied to one customer or a broader set of buyers?
- Software and portability: Can customers use the product with their existing tools and systems, and how much adaptation is required? A hardware comparison alone may miss the importance of the software ecosystem.
- Deployment economics: For the relevant workload, what evidence supports performance, cost, and energy-efficiency claims? Is the chip flexible enough for the customer’s needs, or is its advantage limited to a specific task?
- Revenue quality: Separate reported AI-related revenue from company projections, analyst forecasts, and total-addressable-market estimates. Check the period and definition for each figure.
- Business and investment risk: Examine customer concentration, execution, competition, and valuation. The cited commentary does not provide current comparable valuation data, so it cannot establish whether any stock is attractively priced.
What can be concluded from the available evidence
There are distinct AI-infrastructure investment theses: Nvidia is associated with GPUs and CUDA; AMD’s cited case centers on inference GPUs and data-center CPUs; Broadcom’s centers on custom ASICs and networking. Those differences can help frame further analysis, but they do not reveal which company the headline’s unnamed “under-the-radar” stock is supposed to be. Until that identity and its current business evidence are established, there is no sound basis to name a winner or make a stock-specific recommendation.
Quick Recap
Best Value
- DEEPX DX-M1M NPU: Powered by the DEEPX DX-M1M neural processing unit, purpose-built for efficient on-device AI inference workloads.
- COMPACT M.2 2242 FORM FACTOR: Fits the standard M.2 2242 slot, making it easy to integrate into embedded systems, edge devices, and compact computing platforms.
- EDGE AI ACCELERATION: Designed to accelerate deep learning inference at the edge, enabling real-time AI applications without relying on cloud connectivity.
- RADXA AICORE MODULE: The Radxa AICore DX-M1M delivers a plug-and-play AI compute solution ideal for robotics, smart cameras, and industrial automation.
- WARRANTY AND ORIGIN: Backed by a 1-year manufacturer warranty and crafted with quality components for reliable long-term performance in demanding environments.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
