Under the Radar: 7 Small-Cap Stocks Behind the AI Infrastructure Boom

Somebody is spending a staggering amount of money trying to build a machine that can think. I am far more interested in the companies that keep the lights on while it does. This is where it gets fun. Anthropic,

Somebody is spending a staggering amount of money trying to build a machine that can think.

I am far more interested in the companies that keep the lights on while it does.

This is where it gets fun.

Anthropic, the company behind Claude, just handed investors a shopping list. It sits inside the registration statement filed ahead of its public offering, tucked in among the warnings about what could go wrong.

That is usually where the good stuff hides.

Management uses the opening pages to tell you what it hopes to accomplish. The risk factors tell you what it must buy, what it cannot get fast enough, and who might have the leverage to charge more.

Read those pages, and the numbers jump out.

Anthropic lists minimum spending commitments of approximately $44.9 billion with Microsoft (MSFT), $114.5 billion with Google parent Alphabet (GOOGL), and $110 billion with Amazon (AMZN).

Add them together, and we get roughly $269 billion in commitments.

Much of that spending is owed regardless of how much capacity the company actually uses, subject to contractual exceptions.

Then it gets better.

The filing describes power and data center arrangements lasting 20 years, with annual price increases and uncapped fuel costs.

It also highlights long delivery times and tight supplies of transformers, switchgear, turbines, generators, and transmission equipment.

In some cases, Anthropic may even pay above-market prices to get equipment sooner.

Think about that for a moment.

When the customer is willing to pay extra to cut the line, the seller is holding something the customer badly needs.

And that is where I want to look for investment opportunities.

Follow the Electrons, Not Just the AI Headlines

The opportunity extends across the entire data center.

Power must be generated, delivered, converted, and distributed.

Computing equipment must be installed.

Heat must be removed.

And somebody has to build the enclosures, frames, piping, and controls that hold the whole operation together.

The spending commitments also reveal something important about the customer’s priorities.

Reserving capacity years in advance means management values access enough to commit enormous sums before knowing exactly how demand will develop.

For suppliers, those commitments can support production planning, capacity expansion, and new investment.

For shareholders, however, the fine print determines how much protection those arrangements actually provide.

I prefer suppliers with several large customers.

A manufacturer of electrical equipment serving competing projects has more ways to win than a developer tied to one facility.

The same goes for a service business with a broad customer base and plenty of repeat work.

And we need to keep one important distinction in mind.

A giant spending commitment announced by one company does not automatically become profit for every business associated with the same industry.

Orders must reach the supplier.

Deliveries must happen.

The customer must pay.

And the supplier must keep enough of that payment after expenses to reward its shareholders.

That is why I start with companies already selling relevant products and reporting actual business activity.

We can study something concrete while cable television keeps booking guests to argue about the future of intelligence.

A clever algorithm cannot negotiate with an overloaded transformer.

I found seven companies worth investigating.

Their market values range from approximately $200 million to a little more than $2 billion, based on late-October quotes.

All seven participate in some part of the broader infrastructure buildout.

None has been confirmed to hold a contract with Anthropic.

Please keep that distinction clear.

These are companies positioned to benefit from broader infrastructure demand, not confirmed direct beneficiaries of Anthropic’s spending commitments.

1. American Superconductor (AMSC): Keeping the Power Grid Running

American Superconductor sounds like a laboratory project that needs another decade and three more rounds of financing.

It is not.

The company already sells power controls and grid systems to utilities and industrial customers.

Its equipment helps customers connect, transmit, and distribute electricity reliably.

That job becomes more valuable every time a giant computing facility plugs into the grid.

A utility must serve the new load while keeping the lights steady for everyone else.

Equipment that helps manage those conditions has real work to do.

And American Superconductor is already reporting growth.

Revenue Growth and Data Center Demand

AMSC reported $94.1 million in revenue for the June quarter, an increase of approximately 30%.

Growth came from both existing operations and the company’s Comtrafo acquisition.

Management also identified data center demand as a driver of orders.

That is the kind of evidence I want to see.

We are not relying entirely on projections about how much electricity AI might eventually consume.

We have a business already supplying equipment to the market and reporting growing revenue.

At a market value of approximately $1.35 billion, American Superconductor offers a much smaller way to participate in electrical infrastructure spending than the industrial giants dominating the headlines.

The important questions are how much of that growth becomes sustainable profit, how successfully the company integrates its acquisitions, and whether the valuation leaves enough room for shareholders to benefit.

The opportunity is attractive.

But as always, the price we pay matters.

2. Hammond Power Solutions (HMDPF): Supplying the Transformers Everyone Needs

Hammond Power Solutions takes us directly to one of the equipment shortages Anthropic describes.

The company manufactures dry-type transformers, power quality products, and related electrical equipment.

Its primary listing is in Toronto under HPS.A.

U.S. investors can purchase shares through the OTC symbol HMDPF, which represents foreign ordinary shares rather than an American depositary receipt.

Transformers never get the glamour that semiconductor chips enjoy.

They have the marketing problem of looking like electrical equipment.

Yet a building full of expensive processors needs electricity delivered in a form those machines can use.

And demand for that equipment is growing.

A Backlog That Nearly Doubled

Hammond reported 44.7% second-quarter sales growth.

Its backlog increased 96.9% from a year earlier.

Management pointed to U.S. data center investment, electrification, and power infrastructure spending as important demand drivers.

Those are substantial increases.

And they suggest the company is participating in the infrastructure investment cycle rather than merely talking about it.

Hammond’s market value is approximately $2.12 billion.

I want to see how effectively management uses its expanded production capacity, whether the growing order book translates into stronger earnings, and how much of the favorable outlook is already reflected in the share price.

There is also a practical trading consideration.

The OTC shares can be thinly traded.

Compare the Toronto price with the OTC quote, account for currency conversion, and use limit orders.

A great company does not become a better investment because you volunteered to pay a fat bid-ask spread.

3. Ameresco (AMRC): Building the Power Systems Data Centers Need

Ameresco answers another important question.

Where does the electricity come from?

The company develops and operates energy infrastructure, including systems at customer sites where generation and other resources can be built around specific needs.

That matters when a customer wants computing capacity faster than the local power grid can deliver.

A dependable power solution can determine whether a project gets built at all.

Ameresco can bring together engineering, construction, generation, storage, and ongoing operations.

That creates several ways to participate in a single infrastructure project.

And the company is already winning substantial business.

$1.2 Billion in Data Center Awards

Ameresco’s second-quarter announcement included approximately $1.2 billion in new data center awards.

The company reported five data center projects in awarded backlog.

That is serious commercial activity for a company with a market capitalization of approximately $1.07 billion.

But we need to give those numbers their proper meaning.

An award is not the same thing as a signed construction contract.

Projects must clear development and other milestones before they move into contracted backlog.

Give the company credit for winning the business.

Then watch how quickly those wins become revenue and cash.

I like a developer that can assemble the various pieces of a complicated energy project.

Coordinating engineering, construction, generation, storage, and ongoing operations can create opportunities well beyond the initial installation.

But the financial statements still get a vote.

Project financing, debt, and working capital deserve close attention.

A large project is valuable only if the company can complete the work and earn an attractive return after covering its costs.

Ameresco has a meaningful opportunity as data center operators search for dependable power solutions.

Now management needs to convert that opportunity into profitable execution and stronger cash flow.

4. Mayville Engineering (MEC): The Small-Cap Manufacturer Behind Critical Power Equipment

Mayville Engineering is the kind of company that disappears behind the glamorous names on a supplier list.

And that is exactly why I find it interesting.

MEC manufactures precision metal enclosures, racks, frames, and assemblies for equipment makers.

In its data center and critical power business, those products support power distribution, backup energy, and power management systems.

Somebody has to manufacture the physical equipment surrounding all that computing power.

MEC gets paid to do it.

The recent growth numbers deserve attention.

Data Center Sales Are Growing Rapidly

Second-quarter data center and critical power sales reached approximately $29 million, compared with roughly $5 million a year earlier.

An acquisition contributed significantly to that increase.

But organic growth in the market was approximately 173%.

That second number matters.

Buying another business can make reported growth look spectacular.

Organic growth shows that demand is expanding within the existing operations.

At a market capitalization of approximately $353 million, a growing business line can make a meaningful difference to the entire company.

If the margins are acceptable.

And that condition deserves emphasis.

Execution and Cash Flow Still Need Improvement

New program launch costs weighed on recent results.

Quarterly free cash flow was negative.

Net leverage stood at approximately 2.9 times adjusted EBITDA.

Those are important considerations.

I want to see production become more efficient as the new programs mature.

If higher sales begin producing better margins, stronger earnings, and improved cash flow, the investment case becomes much more interesting.

A company this size does not need to become an industry giant to produce substantial shareholder gains.

But it does need to turn growth into profits and cash.

MEC gives us an opportunity to watch that process develop.

5. Perma-Pipe International (PPIH): The Overlooked Business of Keeping Data Centers Cool

Perma-Pipe supplies another essential part of the infrastructure system.

The company manufactures insulated piping, leak detection systems, and other engineered products for energy, industrial, and infrastructure customers.

That connects it to cooling and fluid handling, which deserve increasing attention as computing facilities multiply.

Powerful computers create a permanent obligation to manage the heat they produce.

Cooling systems must operate reliably around extremely expensive hardware.

Suppliers of those supporting systems can participate in infrastructure spending without ever manufacturing a semiconductor.

And Perma-Pipe is seeing business growth.

Rising Sales and an Expanding Backlog

Second-quarter sales increased 24%.

Backlog reached $142.3 million.

The company also announced its first critical cooling infrastructure award in the Middle East and North Africa region.

Management reported continued backlog growth at its newer Ohio facility.

Those developments give us something tangible to investigate.

Perma-Pipe serves several industries, so we need to determine how much of its business is actually tied to data center cooling before assigning it a fashionable AI valuation.

At a market value of approximately $243 million, successful expansion into additional infrastructure markets could matter a great deal.

The valuation also deserves attention.

Recent figures put the shares at approximately 15.7 times trailing earnings, with enterprise value around 7.9 times EBITDA.

That is a value investor’s kind of setup.

But we still need to review earnings quality, customer collections, project timing, and the cash required to fulfill the backlog.

A modest multiple becomes a bargain when the earnings supporting it are dependable and still growing.

That is what I want to determine with PPIH.

6. Matrix Service (MTRX): An Infrastructure Contractor With Recovery Potential

Matrix Service gives us exposure to the construction work connecting new electricity demand with the energy supply required to meet it.

The company engineers and builds energy infrastructure for utility and power projects, storage facilities, and other industrial applications.

Management has identified data center power generation and delivery as growth opportunities alongside its established LNG and storage businesses.

This solves a practical problem.

Buying equipment is only the beginning.

Projects still require engineering, site preparation, construction, and connections to working infrastructure.

Experienced contractors have plenty of opportunities to pursue as those projects move forward.

And Matrix is small enough for improvements in its business to matter.

An Operating Recovery Opportunity

The company’s market value is approximately $281 million.

Fiscal 2026 revenue increased approximately 14%, although Matrix still reported a small annual loss.

That makes it an operating recovery candidate with exposure to infrastructure spending.

The potential is interesting.

But contract terms are at the heart of the analysis.

A contractor can report an impressive backlog and still disappoint shareholders if costs exceed estimates.

I want to see better execution, sensible bidding, and stronger cash generation as projects progress.

When a company starts from a weak earnings base, even a relatively small improvement in margins can make a meaningful difference.

That is the opportunity here.

We can investigate the potential for recovery without assuming every announced energy project will arrive on schedule.

7. TSS (TSSI): Putting AI Computing Infrastructure to Work

TSS takes us inside the computing facility itself.

The company integrates and deploys AI and other high-performance computing infrastructure.

It turns equipment into systems customers can actually use.

That is a different role from supplying electrical components, piping, or power generation projects.

Complex computing equipment still needs to be configured, integrated, and deployed.

A company that performs that work efficiently has a real opportunity as customers expand their computing capacity.

And TSS is already reporting growth in the business.

Systems Integration Revenue Climbs 46%

Second-quarter systems integration revenue increased 46%.

The company also announced approximately $17 million in investment to support work on the next generation of AI infrastructure.

With a market capitalization of approximately $211 million, continued growth in a successful service line could have a substantial impact on the company’s overall value.

The important questions are whether that growth can continue, how much profit the company earns from the work, and whether the additional investment produces attractive returns.

There is an opportunity here.

But we still need to examine the economics rather than assuming that every company associated with AI infrastructure will become enormously profitable.

TSS gives us another smaller business worth investigating as the buildout continues.

Seven Small Companies, One Enormous Infrastructure Opportunity

These seven businesses offer different ways to participate in the infrastructure spending described in Anthropic’s filing.

American Superconductor (AMSC) helps utilities and industrial customers manage electricity delivery and grid reliability.

Hammond Power Solutions (HMDPF) manufactures transformers and electrical equipment needed to distribute power.

Ameresco (AMRC) develops energy solutions and infrastructure projects, including work associated with data center power demand.

Mayville Engineering (MEC) produces the enclosures, frames, and other physical components supporting critical power equipment.

Perma-Pipe International (PPIH) supplies engineered piping and detection systems, with an emerging opportunity in critical cooling infrastructure.

Matrix Service (MTRX) engineers and constructs energy infrastructure needed to connect new demand with supply.

TSS (TSSI) integrates and deploys computing systems inside data centers.

The variety is part of the appeal.

We can compare business models, balance sheets, and valuations instead of treating every company with an AI label as interchangeable.

Some of these businesses are growing rapidly.

Others have opportunities to improve operations or recover from weak earnings.

Several are small enough that a successful new business line could meaningfully change their financial results.

But none deserves a free pass simply because AI infrastructure spending is expanding.

A growing market attracts new competitors and additional capacity.

Customers postpone projects.

Suppliers overexpand.

And paying too much for a good business remains one of the easiest ways to turn a promising investment idea into a disappointing result.

Follow the Orders, Earnings, and Cash Flow

Anthropic’s filing provides evidence of where pressure is building in the infrastructure market.

The company is making enormous commitments to obtain the computing capacity and power resources it expects to need.

And its disclosures highlight the equipment shortages, delivery delays, and other practical problems that come with building infrastructure on this scale.

That creates potential opportunities for smaller suppliers.

But we need to follow the money all the way through.

An industry forecast is not an order.

An order is not revenue.

Revenue is not profit.

And profit does not always translate into cash for shareholders.

The next step is finding suppliers whose orders, earnings, and cash flow support the story.

That is where the real investment opportunities will be found.

The technology experts can argue all they like about which AI model will end up smartest.

**I would rather own smaller businesses that earn more as the facilities get built, the equipment gets installed, and the e