$632M EV With ~4% Downside

🌾 Welcome to StableBread’s Newsletter!

A few weeks ago I wrote about Boost Run, the GPU cloud infrastructure company merging with Willow Lane Acquisition Corp (WLAC).

Since then the stock has run to ~$21, and more importantly, shareholders approved the BCA with zero redemptions. That's rare for SPACs.

I was alerted to a similar potentially asymmetric setup. This post walks through it, with a direct comparison to WLAC/WLACW at the end.

D. Boral ARC Acquisition I Corp. (BCAR/BCARW) is a $280M SPAC that priced its IPO on July 31, 2025 and has traded between $9.88 and $10.70 ever since.

It sits at ~$10.70 today, above the $10.17 trust floor reported at 12/31/2025. Interest has been accruing since, so today's actual redemption value is closer to ~$10.30.

The target is Exascale Labs, a private AI infrastructure operator that signed a definitive merger agreement with BCAR on January 11, 2026.

Pre-money equity value is $500M and the implied pro forma enterprise value is $632M. Post-close, the combined entity will trade on Nasdaq under XLAB.

What's unusual about this one is that Exascale already has audited GAAP financials.

FY2025 revenue was $7.0M, up 5.3x from FY2024. H1 FY2026 (six months ended December 31, 2025) was $6.8M, almost matching all of FY2025 in half the time.

Annual gross margin expanded from 4.2% to 15.8%. The H1 FY2026 operating loss shrank 12% y/y even as the business kept scaling.

So this isn't a pre-revenue concept SPAC. It's a small business inflecting fast, packaged through a SPAC where the trust still anchors the floor. At ~$10.30 today, that's the downside. The operating story sets the upside.

Pre-money equity value at Exascale is $500M, payable in 50,000,000 newly issued combined-company shares at $10.00 each.

In a zero-redemption scenario, $280M of trust cash transfers to Exascale's balance sheet. Total sources and uses come in at $780M. The implied pro forma equity value is $912M at $10/share, and net of trust cash, the implied enterprise value is $632M.

Exascale rollover holders take 54.8% of the post-close equity in zero redemption, with effective voting control via Class B shares carrying 20 votes each. BCAR sponsor and non-public Class A take 14.5%. BCAR public shareholders take 30.7%.

In a 100% redemption scenario, no trust cash transfers, sources and uses fall to $500M, and the implied EV stays at $632M because equity compresses dollar-for-dollar with the trust outflow. Additionally, Exascale jumps to 79.1% and the public goes to zero.

Expected close is the second half of 2026, originally guided to Q2. The 18-month combination window from BCAR's August 1, 2025 IPO doesn't expire until February 1, 2027, with a three-month sponsor extension on top.

Here’s BCAR's existing capital stack:

The four largest public holders per Schedule 13G filings are Harraden Circle Investments (9.93%), Polar Asset Management Partners (7.30%), Glazer Capital (6.84%), and Linden Capital (5.30%).

Combined that's ~31% of the public float, all SPAC arbitrage shops. Their economics will drive redemption decisions at the vote, not retail conviction.

What's not yet quantified is the SAFE conversion math. Exascale raised ~$14.1M across five SAFE rounds from June 2022 through March 2026, with valuation caps stepping from $25M to $500M.

The SAFEs convert into combined-company Class A at close, but the exact share count is not in the 8-K or the deck. That's an S-4 disclosure.

The deal also explicitly assumes no PIPE (private investment in public equity) financing in either pro forma scenario. PIPEs are private placements to institutional investors, used to backstop redemptions and add cash at close. Without one, BCAR's post-close cash depends entirely on the redemption percentage.

Three business lines run on top of an asset-light deployment model that uses partner data centers in 10 countries. Only the GPU-as-a-Service line is generating meaningful revenue today.

Exascale's GaaS product runs under the Exabits brand and sells reserved and on-demand bare-metal GPU capacity. The current fleet covers NVIDIA GB200, H200, H100, and RTX Pro 6000.

Use cases span large-scale model training, fine-tuning and post-training, production inference and AI agents, and scientific computing.

The differentiated piece is the deployment model. Exabits aggregates GPU capacity across data center partners rather than owning facilities directly.

Per the deck, customers are located in 12 countries. GPU utilization is running at 95.5%, which is high enough to suggest partner capacity isn't sitting idle and is being matched to demand reasonably well.

This is the line generating the $7.0M of FY2025 revenue and the $6.8M H1 FY2026 number. It's also the source of the AI Nova MOU (more on this later).

Exascale designs and ships prefabricated, containerized data center modules at up to 1MW+ per container with rack densities up to 150kW. Cooling is air or liquid, modules are Tier 3-aligned.

The pitch on speed is ~6 months from order to commissioning vs. ~36 months for conventional construction. The pitch on cost is up to 1/3 of a conventional build at the same scale.

The leadership team has been working on modular data center infrastructure since 2019 through prior roles.

Operational services tied to this line generate revenue today (GPU cluster management, monitoring, optimization), but the full MDC packages are positioned as customer-deployment offerings rather than current contributors.

This is the pre-revenue piece and arguably the most differentiated.

AI racks are pulling toward 1MW of power, far beyond what the standard data center power chain was built for. That chain runs grid power through a transformer, a UPS for backup, then distributes as low-voltage AC to each server, with multiple conversion stages bleeding energy as heat at every step.

Exascale's pitch is a power platform that converts medium-voltage grid power directly to 800V DC using Silicon Carbide-based solid-state transformers.

The targeted specs are up to 90.8% end-to-end efficiency vs. ~85.8% for traditional AC, up to 75% faster install, up to 90% copper reduction, and up to 65% smaller electrical footprint.

Management's own footnote labels this offering "ready for commercial engagement but has not generated revenue to date." So this is a forward catalyst tied to whether the company can convert engineering claims into real installs.

Exascale serves 22 enterprise customers across 12 countries with a 92% retention rate from FY2024 to FY2025.

The disclosed roster includes MIT, Hankuk University of Foreign Studies, Lepton.ai (acquired by NVIDIA in 2025), Nebula Block, Near Protocol, FlowGPT, and Colossyan.

Average revenue per customer on an annualized H1 FY2026 basis is ~$300K.

The most concrete contracted datapoint is the December 2, 2025 MOU with AI Nova PTE LTD. The base contract calls for ~$53M of revenue over three years (~$17.7M annualized) running on Nvidia Blackwell GPUs.

Optionality runs up to 10,000 GPUs and >$600M of revenue over five years if AI Nova draws all expansion options.

That's a large deal relative to the current revenue base. $17.7M of annualized revenue from one MOU is meaningful concentration on a $13.6M run-rate business.

The CEO and CFO is the same person, Hoansoo Lee. He has a PhD in Business Economics from HBS and previously served as Staff Economist for Entrepreneurship and Innovation Policy at the White House CEA under Obama.

He's also founder and managing director of HSL Capital Management, a multi-strategy hedge fund. Lee co-founded Exascale in June 2022 and has run it since.

Technical execution sits with three operators:

The Bitmain background, combined with pre-deal investors that include Hack VC, Stanford Blockchain Builders Fund, and Bloccelerate, points to a former-crypto-infrastructure pivot rather than an enterprise-native build.

Here’s the post-close board:

Lead Director Prof. David Card is a Nobel Laureate in Economics (2021) at UC Berkeley, Compensation Committee chair Prof. Shachar Kariv chairs the Berkeley Economics department, and Wenying Jia (chairperson) is an early Exascale angel investor with digital assets and GPU infrastructure experience.

Three of four directors are economists, including the CEO, which is heavy academic weighting for an AI infrastructure SPAC.

Whether it's the right governance signal depends on how much weight you place on academic credibility versus operating experience.

The financials are where the $632M EV either gets defended or doesn't.