The $10.5 Billion Escape Hatch: Firmus, the Miner Exodus, and the Valuation That Has No Witnesses

A Bitcoin miner just raised $2 billion to stop being a Bitcoin miner. The market is calling it transformation. I'm calling it an evacuation — and the $10.5 billion valuation attached to that escape is the most expensive leap of faith I've seen since Terra's algorithmic feedback loop promised to print stability from nothing.

Here's what we actually know: Firmus, a company that built its existence on SHA-256 hashrate, has completed a $2 billion capital raise and now carries a post-money valuation of $10.5 billion as it pivots to AI infrastructure. Sustainable energy is the stated theme. Asia-Pacific expansion is the stated geography. Everything else — the investors behind the round, the capital structure, the GPU specifications, the customer contracts, the leadership team — is a void.

In 26 years of watching this industry — after Tezos taught me to read whitepapers instead of press releases, after Compound taught me to map protocol interdependencies before the cascading liquidation hits — I've learned that voids in disclosures are rarely accidents. They are either negotiations still in flight, or details that wouldn't survive scrutiny.

The miner-to-AI pivot is not new. It has been the industry's favorite escape narrative since late 2022, when Core Scientific filed for bankruptcy and emerged clutching AI hosting contracts. Hut 8 repositioned its entire brand around digital infrastructure. Hive renamed itself Hive Digital Technologies. Iris Energy started buying NVIDIA GPUs and publishing its data center specs. The playbook standardized itself: take the substations, the cooling towers, the industrial real estate, the power purchase agreements — everything built to convert electricity into hashes — and retrofit it to rent compute to AI labs.

By 2024, CoreWeave had signed massive hosting deals with Core Scientific and other miners, proving miners could at least serve as landlords for GPU clouds. Public market reactions followed a Pavlovian pattern: mention "AI" in an earnings call, watch the stock jump. The narrative became self-reinforcing. Miners are AI infrastructure companies in disguise. Their power access is the moat. Their industrial real estate is the fortress.

Firmus is the most extreme expression of that narrative to date. A $2 billion raise for a company that, by all public evidence, is still in the planning stages of transformation. A $10.5 billion valuation that places it above nearly every publicly traded mining company in the United States.

Let me put that in perspective. Core Scientific — with actual AI hosting contracts and actual revenue — sits at a $4 to $5 billion market cap. Hut 8 hovers in the $3 to $5 billion range. Iris Energy, around $3 to $4 billion. CoreWeave, the pure-play GPU cloud darling with billions in Microsoft and Meta contracts, commands roughly $35 billion after its IPO. Firmus, with zero disclosed customers, zero disclosed GPU inventory, zero disclosed facilities, and zero disclosed leadership team, carries a $10.5 billion price tag.

The ledger remembers what the hype forgot.

The Technical Reality: This Is Not Innovation, This Is Asset Redeployment

Let me be forensic about what is actually happening on the technical level, because the narrative frame matters less than the physical reality.

A Bitcoin mining facility and an AI data center share maybe 40 percent of their DNA. The power distribution infrastructure — substations, transformers, switchgear — that transfers cleanly. The industrial real estate transfers. The physical security fencing, the access controls, the basic cooling systems, those transfer. What does not transfer: the compute architecture.

ASIC miners solve one equation, repeatedly, forever. They are single-purpose silicon with no memory hierarchy to speak of, no interconnect fabric, no job scheduling complexity. AI infrastructure runs on GPU clusters connected by high-bandwidth, low-latency networks — InfiniBand or RoCE — with sophisticated schedulers, storage tiers that handle petabytes of training data, and liquid cooling systems that ASIC farms never needed at scale.

This is the part that gets lost in the "miner-to-AI" enthusiasm: Bitcoin mining is embarrassingly parallel. A thousand ASICs work independently, each solving hashes without needing to communicate with a peer. AI training is the computational opposite. Thousands of GPUs must synchronize parameters, exchange gradients, and maintain a coherent state across the entire cluster. The networking requirements are a different universe. A mining facility can be a barn full of noisy machines. An AI data center is a supercomputer with a building wrapped around it.

During my early audits — I spent six weeks reverse-engineering Tezos's self-amending governance model in 2017 while the rest of the media chased token prices — I learned to distinguish between an evolutionary upgrade and a categorical shift. This is categorical. The electricity-to-compute conversion remains, but the engineering complexity multiplies by an order of magnitude. Miners know power. They do not automatically know RDMA fabric tuning, GPU thermal management under sustained load, or the operational discipline required to keep a 10,000-GPU cluster running at 95 percent utilization.

So when Firmus says it is becoming an AI infrastructure company, what it is really saying is: we own the land, the power, and the building shells. We will buy the GPUs, hire the experts, and figure out the rest.

That is not a technology moat. That is a real estate play with a compute lease attached. And it is being priced like a semiconductor breakthrough.

Estimating the Hidden GPU Commitment

Numbers tell us what press releases hide. A $10.5 billion valuation demands revenue expectations. For an infrastructure business at scale, the market typically prices AI data center companies at 20 to 40 times forward EBITDA. If Firmus is being valued on that kind of multiple, the implied EBITDA expectation is roughly $260 million to $525 million annually.

To generate $260 million in EBITDA from GPU infrastructure, you need serious revenue. At prevailing prices for H100-class compute — roughly $2.50 to $4.00 per GPU-hour — and assuming a 70 percent utilization rate, a single thousand-GPU cluster generates something like $15 to $25 million in annual revenue. Gross margins in the AI cloud business run 40 to 60 percent before depreciation. And depreciation is brutal: GPUs are typically depreciated over three to five years, meaning the headline EBITDA number hides the replacement capital cycle.

This implies Firmus needs at least 10,000 to 20,000 high-end GPUs deployed and contracted to justify its valuation — and that is before accounting for how the $2 billion raise is structured. If that raise includes debt, a significant portion of the revenue goes to interest payments before equity holders see anything.

My 2024 ETF coverage forced me to interview three major custodians about their proof-of-reserves methodologies. I learned that the gap between what balance sheets claim and what is physically verifiable is often a canyon. The same discipline applies here: until Firmus discloses its GPU supply agreements — how many units, which architecture, whether H100, H200, or the newer Blackwell generation — the $10.5 billion valuation is a thermometer with no temperature.

What the $2 Billion Actually Buys

Capital efficiency is where the miner-to-AI thesis faces its harshest interrogation. Let me build a back-of-envelope model based on my years tracking infrastructure builds.

A 20-megawatt AI data center retrofit of an existing mining facility costs roughly $150 to $250 million, including liquid cooling retrofits, networking fabric, electrical upgrades, and building modifications. A 20-megawatt facility houses roughly 4,000 to 6,000 H100-class GPUs, depending on power density. So $2 billion, deployed across such retrofits, builds about 160 to 240 megawatts of AI infrastructure. That is a reasonable first phase — on paper.

But here is the catch: the GPUs themselves consume the bulk of the capital. H100s reached prices of $25,000 to $30,000 per unit at peak demand. A 5,000-GPU cluster is $125 to $150 million just in silicon. And the GPU supply chain has its own physics: NVIDIA's allocation queues run 12 to 18 months for large orders, and export controls add a layer of political risk — especially for any compute destined for Asia-Pacific markets.

I documented this dynamic in my coverage of the 2022 mining credit crisis, when capital flight from miners accelerated during the bear market, amplifying the drawdown spiral. The Firmus transition is different: it is happening during an AI hype cycle where compute capital is flowing freely. That is precisely when mining infrastructure gets cannibalized — not by bankruptcy, but by opportunity cost. The miners are leaving the network for something with a better yield.

A mining company can order 10,000 ASIC miners from Bitmain and receive them in quarterly batches. A company ordering 10,000 H100s is competing for allocation with hyperscalers that buy in hundred-thousand-unit tranches. The question isn't whether Firmus can buy GPUs. The question is whether it can buy enough GPUs, fast enough, at prices that leave room for profit, and secure the customers to pay for them — all while competing against entities with deeper balance sheets, longer customer relationships, and proven operational track records.

Comparative Crisis Mapping: The Peer Set

Let me map the competitive terrain, because this is where the miner-turned-AI story gets genuinely uncomfortable for Firmus.

CoreWeave is the 800-pound gorilla. It started as a crypto mining operation — ironically, Ethereum miners — and pivoted to GPU cloud before the crypto winter arrived. It secured Microsoft as a customer. It got NVIDIA as an investor and supplier. It went public at a valuation that reflects years of operating experience, not just a capital raise and a press release. CoreWeave's advantage is not power access; it is the operational layer: the software stack, the customer relationships, the support infrastructure, the track record with enterprise AI workloads. Its $350 billion market cap makes Firmus's $10.5 billion look modest.

Core Scientific's AI hosting deals with CoreWeave effectively make it a landlord to the leader. The revenue is contractual, disclosed, and verifiable in quarterly filings. You can model its economics. You can see the contracts. It has an actual business.

Hut 8 and Iris Energy have publicly disclosed their GPU purchases, their data center specifications, and their AI revenue streams. Their claims are auditable. Hut 8 has real customers for its cloud services. Iris Energy has actual NVIDIA hardware installed and running.

Firmus has disclosed none of that. What separates a $10.5 billion valuation from a $4 billion one for Core Scientific isn't size — it's the credibility gap. Core Scientific's numbers are audited. Hut 8's GPU fleet is visible in SEC filings. Firmus's entire valuation rests on a single claim: we raised money, and we are going to build AI stuff.

We build on sand, then pretend it's bedrock.

The Energy and Asia-Pacific Play: Signal or Distraction?

The sustainable energy angle and Asia-Pacific expansion deserve deeper scrutiny. These are not just marketing buzzwords. They are strategic positioning signals.

Asia-Pacific is where the AI infrastructure deficit is most acute. Japan, South Korea, and Southeast Asian markets have growing AI ambitions but limited domestic GPU capacity. Singapore is constrained by land and power costs. The region's hyperscalers rely heavily on imported cloud capacity from the US. If Firmus is building power-heavy AI infrastructure in APAC — backed by sustainable energy — it is targeting the region with the most acute supply-demand mismatch. That is a real thesis, and I have seen it play out in alternative asset investing: energy access in the right geography is the single most binding constraint on AI infrastructure growth globally.

But "sustainable energy" is doing heavy lifting in this narrative. For a former Bitcoin miner, the ESG framing also functions as reputational purification. Mining accumulated regulatory baggage, environmental criticism, and a persistent association with carbon emissions and energy waste. The word "sustainable" in the press release is not just describing the power source; it is laundering the company's identity for institutional investors that refuse to touch crypto-adjacent assets.

I have covered enough ESG-themed crypto projects to recognize the pattern: sustainability is frequently a compliance costume rather than an operational commitment. Whether Firmus's power contracts actually deliver carbon-free energy — and at what price premium — remains undisclosed. In the absence of a PPA disclosure or a named energy partner, the claim is marketing, not infrastructure.

The Opacity Problem: $10.5 Billion Without a Witness

In my line of work, I have learned to rank information gaps by their severity. Missing financial models can be forgiven if customers are named. Missing customer names can be forgiven if the technology is proven. Missing the technology can be forgiven if the team has a track record.

Firmus has none of those anchors.

We do not know who invested. Is it sovereign wealth money from the Middle East, which would signal strategic patience and possibly captive demand for AI compute? Is it high-yield debt, which would load the balance sheet with obligations that compound execution risk? Is it convertible notes, which would cap the equity upside while the company builds? The $2 billion raise could be any of those structures, and each implies a radically different risk profile.

This is not a minor detail. The difference between equity and debt at $2 billion scale is the difference between a partner and a landlord with eviction powers. Debt at 10 to 15 percent interest in a high-rate environment can consume half the EBITDA an infrastructure business generates in its early years. If the round is debt, the $10.5 billion valuation may be inflated by leverage — a pile of debt on top of a startup that doesn't yet have a revenue line. The optimists call it expansion. The pessimists call it a time bomb with a delayed fuse.

The governance picture is worse. A company raising $2 billion should be able to name its founder, its CEO, its board. The complete absence of team information from the public record should worry anyone who has watched a promising project die from execution failure. AI infrastructure is not a solo sport. It requires deep operational expertise in GPU cluster management, workload optimization, and enterprise sales. Mining teams, however experienced, are not automatically qualified for those roles.

The Bitcoin Security Angle Nobody Is Pricing

Here is the piece that should alarm anyone watching Bitcoin network security: a miner raising $2 billion to leave mining is a directional signal about mining economics. Even if Firmus continues hashing during the transition — which the "asset reuse" playbook typically involves — the end state is clear. It will not be a net buyer of ASICs. It will not be growing hashrate. It may be selling mining equipment into a secondary market as it liquidates its old business.

Other miners are watching. If the Firmus bet works, more capital follows the same path. The Bitcoin network's industrial floor — the miners who provide price support during bear markets by refusing to sell at a loss — becomes smaller. The institutionalization of mining continues, but with a twist: the most capable operators are focusing on AI, leaving the network to whoever picks up the leftover ASICs.

This is not an immediate threat. Bitcoin's difficulty adjustment and the remaining miners' scale make single-entity exits survivable. But the trend line matters. Every miner that transitions to AI infrastructure is a miner that has, at minimum, decelerated its hashrate growth, and at maximum, committed to selling its mining hardware for the GPU upgrade path.

The long-run question for Bitcoin is not whether Firmus specifically exits mining. The question is whether, at scale, the return on invested capital in AI infrastructure persistently exceeds mining economics. If it does, the hashrate growth curve flattens, the security budget argument gets harder to make, and the "store of value" thesis must increasingly rely on something other than the industrial economics of securing the chain.

The future is a bug report waiting to happen.

The Narrative Cycle: Peak or Wane?

Narrative analysis is my home turf, and I will tell you what the Firmus news tells me about cycle position.

The miner-to-AI story has moved from fringe idea to crowded trade. In 2023, it was pioneering — Core Scientific's AI hosting pivots looked like survival adaptations, and Hive's rebranding was treated as curiosity. By 2024, it was mainstream — every mining CEO had an AI slide in their investor deck, and announcements triggered double-digit stock moves. In 2025, with Firmus raising $2 billion at a $10.5 billion valuation, the narrative is entering what I call the "validation excess" phase. This is where the story has attracted so much capital that the marginal participants are extrapolating from press releases rather than verifiable fundamentals.

We saw this exact arc with the metaverse narrative in late 2021, when Facebook renamed itself Meta and any company that mentioned the metaverse in an earnings call experienced a share price bump. Then reality arrived in the form of revenue reports. The companies that were merely "positioned for the metaverse" found that positioning does not pay salaries.

The same pattern applies to AI infrastructure announcements from mining companies. The "announcement effect" has historically peaked during periods when believers and skeptics disagree most strongly. When the broader public market is repricing every asset with an "AI" label, the incentive to claim AI status — regardless of actual AI operations — becomes irresistible. That creates a selection pressure where the most aggressive narratives attract the most capital, which only raises the stakes on delivery.

Firmus's pivot, with its emphasis on "AI infrastructure company" branding rather than "Bitcoin miner diversifying," is itself a symptom of this cycle. Words matter in capital markets. Calling yourself an AI company signals to a completely different investor base than calling yourself a miner. The headline framing is not stylistic. It is deliberate narrative positioning — a bid to escape the crypto discount and capture the AI premium.

FOMO is just poor risk management in disguise. The question is whether the investors in this round understand they are buying a story, or whether they believe they are buying infrastructure.

The 18 to 24 Month Window

Infrastructure transformation runs on a slower clock than market narratives. My experience tracking protocol launches, token generations, and infrastructure deployments has taught me: the market prices the story on day one, but reality arrives 18 to 24 months later.

For Firmus, the validation timeline runs roughly like this.

Months 0 to 6: assembling the team. Hiring the HPC infrastructure engineers, the network architects, the AI workload optimization specialists. This is a profoundly different skill pool from mining operations. In the 2024 ETF custodial work, I noticed a consistent pattern: successful infrastructure companies systematically over-hire talent during the financing window because the talent market for GPU operations is brutal. If Firmus posts a flurry of senior AI infrastructure hires, that tells you the plan is real. If the team stays silent, the transition is still a press release.

Months 6 to 12: securing GPU supply. Whether Firmus has locked NVIDIA's allocation — and at what terms — will determine whether the $2 billion is actually deployable. The GPU acquisition market is extremely tight. If Firmus is using secondary-market GPUs or struggling to procure next-generation chips, the cost structure changes materially, and the unit economics get worse.

Months 12 to 24: first data center sites live, first customers announced. This is the critical milestone. If Firmus has signed a major AI workload contract — with a name you have heard of — the $10.5 billion valuation starts to acquire a factual basis. Without such a contract, the narrative is just a lease on a building with expensive machinery. CoreWeave built its valuation on named contracts with named customers. Firmus must do the same, or the market will eventually demand it.

Months 24 to 36: revenue recognition. Actual reported revenue, then the market can finally judge whether the unit economics work. AI compute pricing is itself a contested frontier. The market has yet to see a sustained bear case for compute demand, but the supply response — including from every miner that pivoted — could eventually be a factor. If AI compute supply grows faster than demand, utilization rates drop, prices erode, and the entire infrastructure layer gets re-rated.

The ASIC Dump: A Contrarian Signal

Here is the contrarian angle nobody in the capital markets seems focused on.

If Firmus is serious about transitioning out of Bitcoin mining, then at some point it will dispose of its ASIC fleet. Tens of thousands of miners, sold into a secondary market that is already absorbing supply from other transitioning operations.

The short-term impact on Bitcoin network hashrate may be minimal — ASICs find buyers, and global hashrate has historically recovered from capitulation events. But the signaling effect is telling. The balance of productive capital in the industry is shifting away from securing the Bitcoin network and toward serving AI workloads. When the best-capitalized industrial operators leave, the network does not immediately collapse. But it narrows. And a narrower network is a more fragile network, all other things equal.

The secondary market for ASICs also gets distorted. If several large miners transition simultaneously, the supply glut drives down prices for older hardware. That can actually benefit new entrants who can buy cheap machines and start hashing at lower capital costs. But it is not neutral for the industry's evolution. Every secondhand ASIC deployed is a vote for Bitcoin mining as a commodity business with thin margins — precisely why the smart capital is leaving.

Structural Risks: The Full Matrix

Let me be systematic, because casual risk assessment has killed more portfolio value than bear markets ever did.

Supplier concentration risk: If Firmus's entire GPU fleet is NVIDIA H100s, it carries not just technological risk but geopolitical risk. US export controls on advanced chips, particularly for Asia-Pacific deployment, make the supply chain subject to policy shifts. A single regulatory change can delay — or cancel — the delivery timelines that underpin the valuation. The fact that the expansion is specifically in Asia-Pacific makes this risk acute.

Retrofit risk: Converting a mining facility to AI specifications is an engineering exercise with meaningful failure modes. The liquid cooling retrofits, the network fabric upgrades, the power distribution changes — all must be executed without a commodity revenue stream to cushion construction delays. Mining equipment can be operational in weeks. GPU data centers take 18 to 24 months. During that window, the capital is deployed but not generating returns, and every month of delay is a month of carrying costs.

Counterparty risk: If the $2 billion is partially debt, the debt holders have priority claims on future cash flows. Infrastructure debt in this sector typically carries covenant structures that can force asset sales or equity dilution under underperformance. Given that no capital structure details are public, this is not a theoretical concern. It is an unknown with material consequences.

Utilization risk: The AI compute market is rapidly adding supply. CoreWeave, hyperscalers, and existing data center operators are all expanding. Anyone who claims to know what GPU utilization rates will look like in 2026 is lying. The demand for AI workloads is real — I have seen enough enterprise adoption data to believe that — but the elasticity of demand at higher price points, and the pricing power of infrastructure providers, remains unproven at scale. For new entrants like Firmus, they will be entering the market as the supply curve shifts outward.

Competition risk: The incumbents have years of operational experience and established customer relationships. A new entrant needs to win deals against companies that already have proven track records. The AI infrastructure sales cycle is long. The qualification process is brutal. The customers would be risking their own training and inference workloads on Firmus's operational competence. Why would they choose an unproven miner over CoreWeave or hyperscalers with track records? The answer might be price, or geographic proximity, or power availability. But none of those advantages have been disclosed.

The Transparency Ledger

In my audit of Terra's algorithmic stablecoin — a line-by-line teardown I published before the collapse, tracing the anchor protocol's yield spiral and proving the math was unsound — I identified a key principle: the complexity of a financial structure is inversely proportional to the transparency of its operators. The more intricate the story, the less willing the protagonists are to open the books.

Firmus's story is simple: miner plus power plus GPUs equals AI infrastructure. And yet the disclosure surface is minimal. Not one customer. Not one site. Not one GPU count. Not one named investor. Not one member of the executive team. For a $10.5 billion company, that is not privacy; that is a vacuum.

Let me state the problem plainly: a $10.5 billion valuation is a claim about future cash flows. Cash flows that depend on customers, technology, execution, and capital costs. Every input into that valuation is currently unverifiable. The price is not a function of information. It is a function of narrative physics.

Alpha is silent until the chart screams. In this case, the chart has not even been drawn yet.

The Opportunity Cost for Investors

For crypto market participants, the Firmus transition story has an immediate implication that the capital markets are treating as transformative: the rising floor on miner valuations. Every miner with genuine power assets in constrained geographies now has an implied "alternative use" optionality. If Firmus can raise $2 billion at $10.5 billion to pivot, so can Hut 8. Or Iris Energy. Or any other miner with discloseable assets.

This is a genuine repricing force — for public equities. The translation to crypto token prices is more ambiguous. Mining tokens, if they exist, do not automatically capture AI infrastructure value. The equity holders own the AI upside. Token holders still hold exposure to hashrate and network participation, which the pivot devalues rather than enhances.

I have seen this play in reverse during the 2021 bull market, when token prices rose on name changes and partnership announcements without underlying protocol development. The pattern was always the same: the narrative premium preceded revenue by months — and sometimes the revenue never came.

What to Watch

If you want to test the Firmus thesis without waiting 36 months for financial reports, watch these signals.

First, who invested. If a major strategic investor — a large technology company, a cloud provider, or a sovereign fund with regional infrastructure ambitions — is confirmed, that is a significant vote of confidence. It would also explain the valuation. A sovereign fund might pay for geopolitical positioning rather than pure financial return. That would change the analysis entirely. Without that disclosure, the capital could be anything, including expensive debt.

Second, GPU procurement disclosure. NVIDIA's earnings calls are a useful public source. When a company like Firmus enters the order book, there are ripples. Watch for allocation mentions, supply chain reports, or any clue that establishes the GPU purchase. If the company cannot secure supply, the $2 billion has no productive destination.

Third, the Intel or AMD option. If Firmus diversifies GPU procurement beyond NVIDIA, that tells you about both supply constraints and cost structure. Multi-vendor procurement signals a sophisticated infrastructure operator navigating constraints. Single-vendor dependence signals a company that took whatever it could get.

Fourth, the first data center opening. The transition from planning to operation, visible through construction announcements, permits, or progress reports. Job postings for site managers, equipment orders, hiring sprees. These are the physical fingerprints of a real build-out.

Fifth, the ASIC liquidation. If Firmus begins selling mining hardware in volume, that corroborates the transition's depth — and signals secondary-market supply for other miners. The secondary market for ASICs will tell you whether this is a genuine exit or a temporary diversification claim.

The Crypto Market Relevance

Let me close the loop on what a crypto-focused readership should extract from this.

It is tempting to see Firmus's transition as evidence of mining's death spiral, but the reality is more subtle. Mining is not dying. Mining profitability is being diluted by capital substitution. The industry no longer has a monopoly on the industrial capabilities — power procurement, facility operations, compute infrastructure — that Bitcoin needs. AI offers a higher return on those same capabilities.

During the 2017 ICO mania, I learned that infrastructure networks only survive when the marginal participant still has a reason to participate. Bitcoin's security model depends on a healthy, competitive mining industry. When the best-capitalized industrial operators leave, the network does not collapse overnight, but the composition of its security providers shifts. The remaining miners may have lower costs or weaker balance sheets or a stronger ideological commitment to decentralized finance. Whatever the profile, the network's margin of safety changes.

Maybe that is the actual story here. Not one company's pivot, but the signal it carries about the industry's center of gravity. We do not often ask why capital leaves. We should. The answer tells us more about the future than any press release ever will.

The future is never evenly distributed. Neither is the pain of bad forecasts.

The Takeaway: Scrutiny Is a Feature, Not a Bug

One last point worth making. The absence of information in the Firmus case is not necessarily evidence of fraud. It may simply indicate a company at the earliest stages of its transformation, raising capital before all details are finalized. Private infrastructure deals routinely close with undisclosed terms, guests, and contracts still being finalized. The absence is not the problem. The valuation is the problem. When the price tags a company at $10.5 billion before the details exist, the market is being asked to fund a bet, not to evaluate a business.

When a $10.5 billion valuation has no witness — when the company's own presentations are not public, when the customers are not named, when the investors are not identified — the market is taking a leap of faith. I have covered enough cycles to know that faith, in this industry, is just another word for unexamined risk.

The ledger remembers what the hype forgot.

In the next 18 to 24 months, when the first Firmus data centers either ignite or stall, the market will get its reckoning. Either way, the data will be revealed. Either way, the value will be proven or lost. Watch the disclosures. Watch the hiring. Watch the GPU deals. The signals are everywhere.

The question is whether the market will read them before the price does.