Bloom Energy ran 1,400% on this thesis. HYLN is 100x earlier, and outputs native 800V DC.

Bloom Energy ran 1,400% on this thesis. HYLN is the same trade, two years earlier.

Power has become the single biggest bottleneck of the AI era. New gas plants quote 2029+ for grid interconnect. Transformers carry 5-year lead times. Hyperscalers are sitting on data center sites they cannot energize. Bloom Energy (BE) figured out one answer — solid oxide fuel cells — and the stock went from $20 to $310 in 12 months. KARNO is a different answer: fuel-agnostic linear generators built on GE Aviation additive-manufacturing tech. Same problem, different solution, far earlier stage. A $637M cap with a $400M+ LOI pipeline, native 800V DC output, and military traction. If commercialization sticks, this is a multi-billion company.

You can build the data center in 18 months. You can't get the power for 5 years.

To understand HYLN you have to understand the constraint it solves. AI training and inference workloads consume electricity at rates that simply did not exist five years ago. A single GPT-5-class training run pulls megawatts continuously for weeks. NVIDIA's latest racks push 130kW each — a 20x increase from typical 6kW racks of 2020. Hyperscalers are racing to build out hundreds of gigawatts of new capacity. The grid cannot deliver it.

The Three Things That Are Actually Broken

Grid interconnect queues are 5+ years deep. PJM, ERCOT, MISO are all reporting interconnect waitlists measured in gigawatts and years, not megawatts and months. Transformer lead times are 4-5 years. The specialized high-voltage transformers needed for data center scale interconnects are made by a handful of suppliers globally — and every single one is sold out. New gas plants are quoting 2029+. Even if you give up on grid power and want to build your own gas plant, the EPC contractors and turbine OEMs (GEV, Siemens) are booked out three+ years.

Why KARNO Is Architecturally Different

Every other answer to this bottleneck — fuel cells, gas turbines, diesel gensets, nuclear SMRs — has a manufacturing process gated by traditional industrial supply chains. Cast metal, stamped plates, hand-assembled rotors. KARNO is built on GE Aviation additive manufacturing — the same 3D-printing process GE uses for fuel nozzles in CFM LEAP jet engines. Output scales with printers, not with foundries. If you want 10x more KARNO cores next year, you buy 10x more Colibrium printers. There is no analog for that in turbine manufacturing. The same dynamic that lets SpaceX iterate on Raptor engines faster than legacy aerospace — additive manufacturing — is what gives KARNO its scalability story.

The 800V DC Architectural Match

This is the part most retail investors miss. In March 2026, NVIDIA published an 800 VDC reference architecture for next-generation AI racks. ABB and NVIDIA, Eaton and NVIDIA, TI and NVIDIA all announced 800V DC reference designs at GTC. The industry is standardizing on 800V DC because it dramatically reduces conversion losses, copper use, and rack footprint compared to legacy AC distribution. KARNO outputs 800V DC natively. Every fuel cell, every gas turbine, every diesel genset outputs AC, which then requires multi-stage conversion to feed an 800V DC rack — losing efficiency at every step. KARNO skips that. If the industry consolidates on 800V DC (and it is), KARNO has a structural efficiency advantage no incumbent can easily match.

U.S. Navy. Air Force. DOD. And the Bloom Energy precedent.

When a $637M company gets contracted by the U.S. Navy, Air Force, and DOD CDAO simultaneously, that's the kind of validation you cannot fabricate. The military buys what works. They also buy what they need to validate independently before it shows up in commercial deployments. Every defense contract HYLN signs is also a free engineering audit by people who know what they're looking at. On the commercial side, the validation stack is being built in real time — UL certification just cleared, VFG 50MW LOI signed, and the broader fuel-cell/AI-power sector is being repriced violently in Bloom Energy's wake.

Tier 1 — Military Validation

Tier 2 — The Bloom Energy Re-Rate Precedent

Bloom Energy was a forgotten clean-tech IPO that traded around $20 a year ago. Then the AI data center power crisis became impossible to ignore. Oracle struck a deal. Then expanded it to 2.8GW. Then AEP/Crusoe/Tallgrass started building gigawatt-scale fuel cell campuses. BE went from $20 to $310 — a 1,400% move in roughly 12 months. Market cap expanded from ~$5B to ~$70B. The same secular driver — hyperscaler desperation for on-site, fast-deploy, fuel-flex power — is what HYLN is now staring directly at. This is not a guarantee. It is a precedent. The path is well-trodden: small clean-power company gets discovered by an AI infrastructure investor, signs a marquee contract, and the multiple re-rates from "speculative junk" to "critical infrastructure."

Tier 3 — The Mindshare Signal

Retail attention rotates fast in this market. HYLN is starting to show up in the same conversations as Bloom Energy, Vistra, Constellation Energy, GE Vernova — the names that have been the AI infrastructure beta winners of 2025-2026. Volume is now 8x relative on key sessions. The stock has more than doubled YTD (+128%) and is testing 52-week highs. Beta is 2.6, meaning it moves ~2.6x the market — both directions. This is what early-stage mindshare looks like before institutional coverage catches up. No analyst has it on a buy list because almost no analyst covers it. By the time they do, the trade will be a different price.

Failed truck SPAC pivoted to AI power. The new company has nothing to do with the old one.

Hyliion Holdings Corp. (NYSE American: HYLN). Founded 2015, went public via SPAC in 2020. Headquartered in Austin, TX. Originally built hybrid powertrain systems for Class 8 trucks (Hypertruck ERX). That business is fully discontinued. The company spent the last several years pivoting entirely into distributed power generation through its KARNO Power Module — a fuel-agnostic linear generator platform. Roughly 145 employees. Market cap ~$637M. Beta 2.6.

What KARNO Actually Is

KARNO is not an internal combustion engine. It is not a fuel cell. It is a linear generator — a category of power generation that uses oscillating pistons (not rotating shafts) to drive electromagnetic induction, producing electricity directly. The thermodynamic cycle is closer to Brayton (gas turbine) than Otto/Diesel (ICE), but the architecture is mechanically simpler than either. The components are built using GE Aviation's additive manufacturing process — the same metal 3D-printing capability GE uses to build commercial jet engine parts. This is the part that matters: production scales with how many printers you operate, not how many factories you build.

The Five Things That Make It Different

The GE Aviation / Colibrium Pipeline

Production capacity is gated by Colibrium additive-manufacturing printers (the company that emerged from GE's metal 3D-printing business). Current printer fleet plus planned additions are projected to support 2026-2028 production scaling. The constraint that bulls cite as the moat — you scale with printers, not foundries — is also the most important operational risk to track. If hyperscaler demand materializes faster than printer adds, HYLN becomes capacity-constrained. If demand is slower than printer adds, HYLN burns cash on unused capacity. Watch the printer fleet count quarter by quarter.

Customer & Market Targets

KARNO's commercial targets span AI data centers (the primary thesis), military systems (current revenue), microgrids and prime backup generation, industrial power for remote oil/gas/mining, and EV charging infrastructure. Management is guiding to ~10 early adopter deployments in 2026, meaningful production scaling in 2027, and accelerated commercial growth into 2028+. The data center pipeline is where the multi-bagger lives. Everything else is supplementary.

Pre-revenue with a long runway. The math works if commercialization lands.

Q1 2026 Results (Reported May 12)

The Pipeline Math

Hyliion currently has ~750 KARNO Cores under non-binding LOI, representing >$400M of potential revenue at current pricing. The VFG Holdings LOI alone is up to 250 cores / 50MW over 5 years. There is also an expected $40-50M in additional U.S. military contracts management is guiding toward in 2026. Even partial conversion is material. If just 30% of LOIs convert to firm orders by end of 2027, that's $120M+ of revenue against a $637M market cap — and the multiple expands as the company crosses the commercial-scale threshold.

The Valuation Math

HYLN is too early-stage for traditional valuation frameworks. There is no meaningful forward P/E (still loss-making). EV/Revenue is uselessly high (50x+) because revenue is just starting. The only framework that matters here is the BE precedent. Bloom Energy traded around $20 with a ~$5B market cap in early 2025 — before the AI data center power thesis was widely understood. It now trades at $310 with a ~$70B market cap. That is a 14x multiple expansion. HYLN at $637M is roughly where BE was when the re-rate began. The bull case isn't "fair value is X" — it's "if commercialization works and AI power demand stays inelastic, the market will eventually pay for HYLN what it has already paid for BE."

The Burn Math

$139M cash. ~$50M annual burn. ~2.8 years of runway at current spend. Year-end 2026 target is ~$100M cash. That means HYLN can comfortably get to end of 2027 without raising. By then, if the ~10 early adopter deployments deliver and commercialization scales as guided, revenue is in the $30-60M range and the question shifts from "can they survive" to "how big can they get." Dilution risk is real but not imminent. The next equity raise will likely be opportunistic (off a higher stock price) rather than necessary (off a cash crunch). If the stock keeps running, expect a raise — and read it as bullish, not bearish.

Analyst Coverage

Sell-side coverage is thin. This is a small-cap energy story still flying mostly under institutional radar. This is a feature, not a bug. Most of the multi-baggers in the AI infrastructure cycle (BE, VST, SMR, AVAV) were under-covered when the trade was best. Coverage arrives after the move. Position before it.

Every other answer is gated. This one scales with printers.

The AI power bottleneck is real and well-funded. There are essentially four categories of solution, and each has a structural constraint that limits how fast it can scale. Understanding why each is constrained is the only way to understand why KARNO is differentiated.

The Four Buckets Of On-Site Power For AI

HYLN vs BE — The Direct Comp

Bloom Energy is the closest comparable — both are on-site fuel-flex power technologies serving the AI data center market. The architectural difference is meaningful. Bloom scales by stacking solid-oxide fuel cell modules at its Sunnyvale facility — capacity is gated by the speed of building, equipping, and ramping fuel cell manufacturing lines. HYLN scales by adding 3D printers — fundamentally a different capex profile. If a hyperscaler decides it wants 1GW of KARNO capacity by 2028, the limit is how many printers HYLN (or a partner) can put online. There's no analog for that in fuel cell manufacturing. BE has the proven track record. HYLN has the scaling architecture.

HYLN vs Gas Turbines (GEV)

GE Vernova's gas turbine order book is sold out through 2028-2029. That's the bull case for GEV (it's traded as such, +400% over the past 18 months) and simultaneously the case for KARNO. Hyperscalers who couldn't get a turbine slot need an alternative — and they have money. The deeper irony: KARNO uses GE Aviation's metal printing technology. So GE benefits either way — directly via GEV turbine sales, indirectly via printer sales to Hyliion. The capital-cycle math actually favors HYLN here because GEV gets paid once per turbine; printer supply to HYLN is recurring.

HYLN vs FCEL / PLUG

FuelCell Energy and Plug Power are the two other publicly-traded fuel cell names that ran with the AI power thesis. Both rallied hard but neither has the same architectural advantage (native 800V DC, additive manufacturing scaling). FCEL's molten carbonate technology has efficiency issues at scale. PLUG has chronic profitability issues and a legacy hydrogen forklift business that drags the model. HYLN's positioning vs these two is cleaner — earlier stage but with better tech and a stronger balance sheet (cash/burn).

Real bull case. Real bear case. Sizing is the only honest answer.

Speculative micro-cap. Sized small. Multi-bagger optionality if execution lands.

Position Structure

This is NOT a core position. This is a high-conviction asymmetric bet. The right way to size it is small enough that the bear case is survivable and large enough that the stretched case is meaningful. 1-2% of portfolio in common stock is the right ballpark for the size of trade John runs. The path to $25+ requires execution that is real but not guaranteed. The path to $1.80 requires execution that is plausible. The dollar-asymmetry only works if position size is small enough that you can sit through a 50% drawdown without flinching.

Options Considerations

Implied volatility is elevated (recent earnings move ~8% expected vs ~30% realized). LEAPs are technically available but premium is rich. Cleaner trade is common stock with mental stop, not options. If you want defined risk, a Jan 2027 $5 call or call debit spread is the right structure — but theta will work against you given the long execution timeline. For a 12-24 month thesis with binary commercial execution risk, equity is the better tool.

Catalyst Calendar

Smallest cap. Earliest stage. Most asymmetric upside if it works.

The Three Comps That Matter

HYLN vs BE: Direct thesis comp. Bloom traded around $20 and ~$5B market cap before the AI data center re-rate began. HYLN is currently at $3.60 and ~$637M market cap — roughly where BE was at the start of the move. If HYLN follows BE's path, the upside is 10-15x. If it doesn't, the floor is the cash balance ($139M / ~$0.78 per share). HYLN vs SMR: Both pre-commercial AI infrastructure power plays. SMR is at $6B vs HYLN's $637M. SMR's first deployments are 2030+; HYLN's first deployments are 2026. HYLN is earlier-stage on cap but later-stage on actual deployment. That's the asymmetry. HYLN vs FCEL: Same market cap range (~$300-637M). FCEL is commercial but sub-scale with molten carbonate tech that has efficiency issues. HYLN is pre-commercial but the tech architecture (additive manufacturing + 800V DC native + fuel-agnostic) is structurally cleaner. If the AI power story rewards architecture over installed base, HYLN wins.

Small position. Long. Sized to survive a 50% drawdown.

HYLN is the trade you put on with size you can stomach losing. Not because the thesis is bad — the thesis is real. Power is the bottleneck of the AI era. KARNO has a genuinely differentiated architecture, military validation, UL certification, a $400M+ LOI pipeline, and a balance sheet that gets to the first commercial deployments without dilution. The BE precedent shows what happens when a clean-power company gets discovered by AI infrastructure capital. But — pre-commercial means execution risk. Beta 2.6 means volatility. A $637M micro-cap means sentiment can swing 20% on a single headline. The right way to play it is asymmetric position sizing, not asymmetric conviction. 1-2% of book. Equity, not options. Sized so the stretched case (10-bagger) matters and the bear case (-50%) doesn't ruin you.