Angel Publishing's 'Homestacks' — The Three Stocks Behind the Nvidia-Backed Data Center Pitch

The Hook

Angel Publishing is selling a $99 special report, no refunds, under its “Angel Ancillary” banner, the arm it uses for one-off special reports that are not attached to any specific newsletter and usually do not name a specific pick-maker. The report is built around a single idea: the giant centralized data center model is hitting a wall, so the future of AI computing is a distributed appliance installed on brand-new homes.

The pitch opens with a claim that new data center construction is running into opposition everywhere, and not just from general AI fear or routine NIMBY complaints. The specific objections, the promo argues, are about how much power and water these facilities consume and how much heat they throw off. The proposed fix is a product Angel calls “Homestacks,” described as “the AI infrastructure Nvidia just bet on to replace Big Tech’s $7 trillion data center empire, rolling out now on brand-new homes in Arizona and Nevada.”

Here is the actual product. It is made by a private San Francisco startup called SPAN, and its real name is XFRA, not “Homestack.” It is an outdoor appliance, about the size of an AC compressor unit, that pairs NVIDIA Blackwell GPUs and memory (roughly $150,000 or more of equipment per unit) with a smart electrical panel and a home battery. SPAN covers the host homeowner’s power and internet bills and installs the battery and panel for free. The first pilot is about 100 units in newly built, build-to-rent communities in Arizona and Nevada, working with homebuilder PulteGroup.

The hook for investors is that you do not have to host one to profit. “Three publicly traded companies are quietly building the supply chain.”

The Big Claim

The promo’s urgency mechanism is a supply and demand gap. “AI demand is doubling every six months,” the copy says, while “giant data center capacity can only grow at the speed permit offices allow, 4-7 years per facility. Those numbers cannot meet.” The conclusion is stark: “Every gigawatt of AI demand above what giant data centers can absorb has to flow through the alternative. There is no third option.”

The investment claim rides on that framing. The three supply-chain companies, Angel argues, are “still being priced by the broader market as if the giant data center build-out is going to continue uninterrupted,” and that pricing window “is scheduled to close when the first 100 Homestacks come online and the financial press starts covering the rollout in earnest.”

There is no specific catalyst date here the way a “buy before August 26” promo has one. The clock is the first 100-unit rollout itself, which is happening now in a handful of Southwest communities.

The Mechanism

The three picks each have a distinct role in the story Angel is telling.

Pick one, the housing play, is nVent Electric (NVT). The teaser logic: every Homestack has to live inside a weatherproof, industrial-grade housing that can protect tightly packed computing hardware, a battery, and a sealed liquid cooling system in 110-degree Arizona heat. The clues point to a company that, “nearly two years before Nvidia announced the Span partnership,” paid $695 million in cash for a business that builds utility-grade outdoor housings. That company’s data center business has been growing about 40% a year and was a billion-dollar segment by 2025. StockGumshoe’s Travis Johnson confirms it is nVent, and it is already a market favorite, up close to 100% over the past year and trading around 25 times forward earnings. We took a closer look at nVent’s enclosure and cooling franchise separately.

Pick two, the comeback play, is Generac (GNRC). The pitch: a complete home energy system needs three layers, battery storage, smart software, and grid controls, and “there is exactly one publicly traded company in America that owns all three layers.” The stock is down about 45% from its high, and the market is “pricing it as a struggling generator maker with a slow software pivot.” Meanwhile, the promo says, Generac has quietly built more than $700 million in data center backup-power backlog, including a non-binding agreement worth up to $600 million with a Big Tech giant. Generac’s home-energy stack is the subject of its own deep dive.

Pick three, the hedge, is Vistra (VST). The teaser describes a publicly traded power company with heavy Texas (ERCOT) exposure sitting directly in the path of surging data center demand, already holding long-term power purchase agreements with two of the five hyperscalers. That is Vistra, which has PPAs with Amazon and Meta and a 20-year deal for half of its Comanche Peak nuclear plant. The “hedge” framing is the clever part: Vistra wins if the Homestack rollout goes well, and it wins just as much if the rollout stalls and centralized data centers keep overwhelming the grid, because “either way, somebody has to provide the power.” We’ve broken down Vistra’s role in the AI power trade elsewhere.

The Real Pick

All three names are real, liquid, exchange-traded companies, and the prices below are live Polygon data from the August 21, 2026 close, which matches the “Buy Price” column in StockGumshoe’s teaser summary.

There is no tease-date price history to measure performance against. StockGumshoe’s Teaser Summary shows “N/A” for both Last Close and % Change on all three, which is itself worth noting: this report carries no performance accountability table, the way a single-stock teaser from a year ago would.

Does the Math Check Out?

The headline claim is the weakest part of the pitch, and it is worth taking it apart number by number.

First, the “$7 trillion data center empire” that Homestacks will supposedly replace. SPAN’s own press release contradicts the promo’s framing. It says XFRA “is not intended to replace centralized data centers, but instead augment them by accelerating capacity growth at the grid edge.” That is a fundamentally different claim than “replace Big Tech’s $7 trillion data center empire.”

Second, the “Nvidia bet on this” language. NVIDIA has agreed to sell SPAN Blackwell GPUs. It has not invested in the project, and it is not discounting the chips. As Johnson puts it, the outcome “is not going to matter at all to NVIDIA’s finances whether or not this test works.” NVIDIA partners with hundreds of companies on experiments like this, including designing GPU packages for data-center-in-space satellites.

Third, the scale gap. Getting XFRA to 1 gigawatt of distributed capacity would mean installing roughly 80,000 units at individual homes or small commercial sites. The actual test is 100 units in a handful of Pulte build-to-rent communities. SPAN has raised around $500 million and taken a $75 million strategic investment from Eaton, but a rollout of that size would require far more capital.

Now put numbers next to the three stocks. nVent at $151.98 and about 25 times forward earnings, with data centers roughly 40% of revenue. Generac at $205.98 and about 20 times forward earnings, with analysts expecting growth to slow to about 20% a year in 2027 and 2028. Vistra at $136.21 and about 13 times forward earnings, with 15-20% earnings growth expected. None of these is a wild valuation, and all three are genuinely exposed to the data center buildout. The problem is that the “Homestack” is not what is driving any of them. Every one of these companies would likely be in worse shape, not better, if centralized data center construction slowed enough to force the industry toward a distributed alternative.

What They Got Right

1. The permitting bottleneck is real. Data center construction genuinely is constrained by 4-7 year timelines, and local opposition over power and water is a legitimate industry concern that has shown up in headlines throughout 2026.
2. The three picks are real supply-chain names with genuine data-center exposure. nVent’s enclosures and cooling, Generac’s backup power, and Vistra’s generation all sit directly in the path of the buildout.
3. The “hedge” framing on Vistra is actually thoughtful. A power producer that wins whether the buildout accelerates or stalls is a coherent position, and Vistra’s two hyperscaler PPAs and the Comanche Peak deal are documented facts.
4. The home-energy angle on Generac is directionally sound. The company has spent years assembling battery, software, and controls capabilities, and a $700 million data-center backup-power backlog is a real number, not a fantasy.
5. The report is honest about being a $99, no-refund, one-time special report, and StockGumshoe’s characterization of it as recycled greatest-hits material is fair.

What They Got Wrong

1. The “Nvidia bet on this” framing is overstated. NVIDIA agreed to sell chips, not to invest, and the promo implies a deeper partnership than exists.
2. The “replace the $7 trillion data center empire” claim contradicts SPAN’s own press release, which explicitly says XFRA is meant to augment, not replace, centralized data centers.
3. None of the three stocks has a meaningful business relationship with SPAN. StockGumshoe found no evidence nVent is a named supplier to SPAN, and Generac and Vistra are not working with the startup at all.
4. The scale math is aspirational. The 80,000-unit, 1-gigawatt roadmap is a white-paper target, while the actual test is 100 units in a few build-to-rent subdivisions.
5. There is no performance accountability. With Last Close and % Change blank in the teaser summary, the report gives readers no way to measure whether the picks have done anything since the pitch began running.

The Verdict

This is a readable, reasonably-priced trio of data-center-adjacent companies wrapped in a fresh fear hook, not a real play on SPAN’s edge-compute appliance. The three picks are legitimate businesses trading at sane multiples, and if the centralized buildout keeps rolling, all three should keep growing. But the specific product the report is named after, the “Homestack,” has essentially nothing to do with their fundamentals, and the promo’s own framing overstates the Nvidia connection and misreads SPAN’s stated purpose.

If you want broad data-center exposure, nVent and Generac are defensible at these prices, and Johnson’s own take is that Vistra is the easiest of the three to buy today given its valuation and his familiarity with it. Just buy them for the data-center buildout thesis, not because you expect a 100-unit pilot in Arizona to reprice them.

This is not financial advice. NewsletterVetter has no position in any stock mentioned.