$1.7B of Contracts in Hand, and the Stock Fell 3.5% That Day

The headline results (unless otherwise noted, all figures from the company's 2026-08-10 8-K earnings release, exhibit EX-99.1, and the same-day press release):

The full-year guidance issued alongside the new contract on August 26 (EX-99.1, 2026-08-26): revenue of $435–445M, Adjusted EBITDA of $105–115M, and capex of $490–510M (excluding acquisitions). The $440M revenue midpoint sits 5.8% above the then-prevailing consensus of $415.8M (four analysts covering, as of 2026-08-26). Sixteen days earlier, on August 10, those same two lines had been $410–420M and $85–95M — revenue raised by $25M, EBITDA raised by $20M.

$25M of extra revenue producing $20M of extra EBITDA is an 80% incremental margin, and that ratio is the one line in the whole release worth stopping on. The press release explains where it comes from: for the new contract, "By leveraging existing portfolio assets, the Community modifications are expected to require less than $15 million of capital investment." Existing lodging assets, refitted for under $15 million. Convert an idle oilfield worker camp into a data center worker camp — same bed, different guest — and the rent more than doubles. You can do this arithmetic yourself from the company's own disclosures: roughly $250M of contract revenue, roughly 1,100 beds, running through August 2030, works out to about $156 per bed per night (our own calculation from the disclosed contract value, bed count and term), against an average daily rate of $72.36 on the HFS–South side over the same period.

So what does the full-year guidance require from the final two quarters? First-half revenue was $158.2M; subtract that from the $440M midpoint and the second half needs $281.8M. First-half Adjusted EBITDA was $28.2M; subtract that from the $110M midpoint and the second half needs $81.8M. Put those together and the implied second-half EBITDA margin is 29.0%, against 17.8% in the first half and 21.3% in Q2 alone. On revenue, the second half needs to average $140.9M per quarter — 65% above Q2's $85.5M. If Q3 lands around $120M, Q4 has to jump to $161.8M. And $161.8M annualized is exactly $647M, which leaves only the tail end of a ramp between there and the $750M 2027 exit run-rate. The curve is internally consistent — provided the beds get delivered on schedule.

The single number to watch in the next quarterly report is WHS average utilized beds. Q2 just cleared 4,000, while contracts signed since January 2026 account for more than 9,000 beds (8-K, 2026-08-10), including two large contracts covering roughly 7,300 beds that need to finish ramping within a year and be complete by mid-2027 (Q2 earnings call, 2026-08-17). Revenue is beds times rate times nights occupied; the rate and the term are written into the contracts, so the only variable that can move is when the beds actually fill.

The demand-side evidence is solid. Since January 2026 the company has signed more than $1.7B of cumulative multi-year contracts (EX-99.1, 2026-08-26), and management says its pipeline of letters of intent "continues to exceed 20,000 beds." CEO Brad Archer's words: "Accelerating demand and continued pipeline conversion are building significant momentum across our WHS segment" (EX-99.1, 2026-08-26). That isn't vague talk about a good industry backdrop — it's already on paper as signed contracts.

A company guiding to $110M of EBITDA at the midpoint plans to spend $500M at the midpoint on capex in the same year. About $123M went out the door in the first half, leaving $370–390M for the second half — against second-half EBITDA that, backing out from guidance, comes to $77–82M. So where does the gap get filled?

The company discloses three sources of funding:

Set those three against the gap and the conclusion is that second-half spending has to lean on the new ABL. One definitional caution here: the CFO's "over $100M" of prepayments sits against a balance sheet line for deferred revenue and customer deposits of $21.3M at June 30 (versus $9.3M at the end of 2025). The two numbers are far apart, which tells you most of that prepayment has already been absorbed into revenue or other accounts. Don't treat "$100M of prepayments" as a pile of cash sitting on the balance sheet waiting to be spent.

On dilution, the market's version of the story needs checking against what was actually disclosed. There were two secondary offerings, one in April 2026 and one in May 2026: April was 7 million shares priced at $14.00, roughly $98M; May was 7 million shares priced at $17.00, roughly $119M (company press release, 2026-05-28). In both cases the sellers were Arrow Holdings S.à r.l. and MFA Global S.à r.l., entities controlled by TDR Capital. The company itself issued no new shares and received no proceeds. Per-share dilution to existing shareholders was zero — Q2 weighted average share count was 100.2M, consistent with the stated share capital. What did change is the shareholder register: a 13D/A shows TDR-affiliated holdings at 47,322,869 shares, or 47.5%, after the May placement. Reading these as "the company raised $200 million" gets it backwards. They were the controlling shareholder selling down — and in the process they pinned two very firm institutional cost bases at $14.00 and $17.00.

On earnings quality, the gap between GAAP and adjusted figures is depreciation and interest. Q2 EBITDA was $18.2M against a $9.0M net loss; the roughly $27M in between is mostly depreciation and amortization on the new fleet plus financing costs. Trailing-twelve-month net income is -$37.68M, or -$0.38 per share (stockanalysis.com, 2026-08-26). This is the normal shape of a heavy-asset build-out phase — every bed being built today starts depreciating on its accounting schedule immediately, while the contract revenue only arrives once the community opens. The real risk is leverage. The CFO's stated target is net leverage "well below three times" on exiting 2027 (Q2 earnings call, 2026-08-17). If annualized EBITDA is $300M at that point, three times implies roughly $900M of net debt — against net debt of only about $34M today. That self-imposed guardrail tells you, in reverse, just how much the balance sheet is expected to swell over the next five quarters.

Upstream is modular housing manufacturing capacity, land, power interconnection, and hospitality labor. Management's framing is blunt: "We've locked up enough line time. We absolutely have the ability and bandwidth to take on more projects, multiple... We don't have an issue at this point with supply" (Q2 earnings call, 2026-08-17). Production slots are already reserved, so the real bottleneck isn't steel or modular boxes — it's capital and construction time. Each 4,000-person community requires more than a hundred million dollars up front and a year of waiting before it generates full revenue.

Downstream there are three types of payers, and they look very different. First, hyperscalers. The North Texas Data Center Hub signed on April 1 serves roughly 4,000 people, carries committed minimum revenue above $550M, runs about five years through Q1 2031, and includes two two-year extension options out to 2035, with net capex of $115–125M — roughly 80% of it landing in 2026 (company press release, 2026-04-01). The West Texas Pecos project announced August 26 serves roughly 1,100 people, is expected to generate about $250M of revenue through August 2030, and requires under $15M of modification spend. The announcement describes this as "another top-five hyperscaler," meaning at least two different giants — but neither release names the customer. You cannot look up who they are, and that is an information gap investors have to accept. Second, government. The Dilley, Texas facility's contract runs through 2030, and the segment's Q2 adjusted gross profit flipped from -$1.1M a year ago to +$6.4M. At the same time, management was explicit: "We're not focused on growing the government segment" — capital is going into WHS. Third, legacy oil and gas customers, where occupancy has fallen from 76% to 70% and volumes are shrinking.

The soft spots in the contract structure need to be seen in three layers. The $550M figure is described as committed minimum revenue — a firm obligation, and the hardest layer. The $250M figure is described as "expected to generate approximately," with the release noting the contract carries "minimum contractual commitments and take-or-pay features" but never disclosing what the minimum commitment actually is — an expected value plus an unquantified floor. And the $1.7B headline is a cumulative contract awards number that blends both definitions together. Easier still to misread is the 2027 figure: "annualized revenue exceeding $750 million and annualized Adjusted EBITDA above $300 million exiting 2027" describes the run-rate at the end of 2027, not the full-year 2027 income statement. Treating $750M as next year's revenue for valuation purposes overstates an entire year systematically.

The most fragile link is the customer's own project schedule. Take-or-pay protects baseline cash flow within the contract term, but it doesn't guarantee that extension options get exercised, and it doesn't produce a second tenant once the contract ends. A camp built in the West Texas desert, custom-fitted for one specific data center site, has a residual value that depends entirely on whatever gets built there next. The second-most fragile link is interest rates. Once the ABL balance climbs above $40M, a cost of funds anchored to a 10-year Treasury yield of 4.64% (FRED, 2026-08-26) starts eating directly into the returns on a heavy-asset build-out.

There's really only one comparable public company: Civeo (CVEO), which also provides worker accommodation and support services at remote work sites. Put the numbers side by side and the directional difference is stark (all data as available on 2026-08-26):

Less than half of Civeo's trailing revenue, a higher full-year EBITDA guide than Civeo, and 4.8 times Civeo's market cap — that's the premium the market assigns to the "AI infrastructure lodging" label over the "oilfield lodging" one. Part of that premium is defensible on margins: TH's Q2 EBITDA margin was 21.3% against Civeo's 13.2%, and TH's WHS segment gross margin is around 53.5%, while Civeo's EBITDA actually declined year over year this quarter on Ontario start-up costs and Australian cost inflation. Same industry: one company climbing a new-customer ramp, the other absorbing costs from old ones.

Civeo is moving in the same direction — it has partnered with a national contractor to develop a 31-acre workforce housing village at Holly Ridge serving Meta's Hyperion data center project (per industry reporting from Bisnow). So this isn't a private lane. But TH's CEO describes the competitive picture as "Most are regional players, some private equity-owned on that side... there is definitely some competition out there," with intensity "not as great as what you might think" (Q2 earnings call, 2026-08-17). Regional operators and PE-owned private firms will struggle to conjure 4,000 beds for a top-five hyperscaler inside a single quarter.

Within the same reporting window, guidance across the specialty commercial services peer group also skewed positive: AMTM raised on 2026-08-10, BKSY and TIC each maintained on 2026-08-06, and DLB gave initial guidance on 2026-07-30. The business models are too different for a numerical comparison, but not one of them guided down in this window — a sign that the overall demand backdrop is favorable.

The price-to-earnings ratio is a useless yardstick here. Trailing EPS is -$0.38, and the forward P/E shows 73.28 (stockanalysis.com, 2026-08-26) — but that 73x reflects new-fleet depreciation pressing down on GAAP earnings, an accounting outcome with little bearing on the cash economics of the business. The right yardstick for a heavy-asset leasing business is EV/EBITDA, cross-checked against contract coverage.

The arithmetic and the results:

What does 15.5x mean in practice? For a leasing company whose asset utilization is still ramping and whose quarterly EBITDA just jumped from $3.5M to $18.2M, that multiple isn't cheap — it already prices in the steep ramp curve for the remaining two quarters of 2026. And 5.7x looks like money lying on the ground, but the cost is in the parentheses: it uses the annualized run-rate at the end of 2027 while pairing it with today's clean balance sheet carrying only $34M of net debt. In the real 2027, the denominator only holds if all 7,300 beds are full, and the numerator rises in step with several hundred million dollars of new debt.

What the market has already paid for at $16.78 is quite specific: a 65% sequential revenue ramp in the second half of 2026, a margin jump from 17.8% to 29.0% over that same half, and two unnamed hyperscalers not rescheduling their projects. What it has not paid for is the 20,000 beds in the pipeline — by the company's own framing, the $750M/$300M 2027 target "assumes no contribution from the Company's commercial pipeline." That's the room left for upside, and it's why this name is worth tracking.

The cycle underneath this business is really the price and capacity cycle for remote-site workforce lodging. For twenty years it tracked the rig count — oil fell, camps emptied. Now it is re-anchoring to the pace of data center construction starts. Three readings help locate where in the cycle we are.

Prices are diverging. On the legacy oil and gas side, volumes are shrinking and prices are flat — HFS–South's Q2 average daily rate of $72.36 is a touch above the prior year's $69.62, but occupancy fell from 76% to 70%, so the rate increase comes from mix, not pricing power. On the new-customer side, pricing is in a different universe: from the disclosed Pecos contract terms — about $250M of revenue, roughly 1,100 beds, running through August 2030 — you get roughly $156 per bed per night, more than double the legacy rate. The North Texas contract, worked from the $550M committed minimum, roughly 4,000 beds and a five-year term, implies about $75 a night, or closer to $96 including the roughly $30M of annual variable revenue mentioned on the earnings call. Within one company, per-bed economics differ by more than 2x across the two customer types — which tells you pricing power on the data center side currently sits with the supplier.

Capacity tightness hasn't hit its limit. The pipeline exceeds 20,000 beds, more than 9,000 are contracted, and 4,000 were actually in use in Q2 — contracted volume is more than double utilized volume, while production line time is already locked up and management is explicit that supply is not a constraint. Supply is neither maxed out nor showing signs of new capacity flooding in and pushing prices down.

Payback period is the most persuasive of the three readings. The North Texas contract carries $115–125M of net capex against a five-year committed minimum of $550M, or roughly $110M of average annual revenue; applied against the WHS segment gross margin of roughly 53.5%, that's about $59M of annual gross profit — a payback of a little over two years. The Pecos contract requires under $15M of modification spend against roughly $62.5M of average annual revenue, a payback measured in months (payback arithmetic is our own calculation). When new capacity in an industry pays for itself inside two years, that industry is still a long way from oversupply.

Put the three readings together and this cycle sits in the early-to-middle part of its upswing: prices rising, supply not saturated, payback extremely fast — and the demand side's capex cycle only entering its second year. There are two leading signals to watch for a top. The first is capex per bed rising and payback stretching from two years out past four — that would mean the good locations and the ready-made assets have been used up, that new contracts require fresh investment rather than conversion, and that marginal returns are collapsing. The second is a turn in demand: hyperscaler capex guidance being cut, or Texas data center construction starts flattening, alongside the company's disclosed pipeline bed count falling below 20,000. Either signal, whichever comes first, would show up one to two quarters ahead of the revenue numbers.

This section covers week-scale signals, relevant to entry and exit timing over the coming weeks. If you plan to hold for a year or more, skip straight to the next section.

Start with the backdrop (all data as of 2026-08-26). Market state is neutral, with a strength reading of 2 (on a scale from -100 to +100). SPY is at 766.01 against a 20-day moving average of 766.64, a gap of -0.08%, while the 20-day line's trend relative to the 50-day is +1.82%. The index is hugging its short-term average — the trend isn't broken, but the momentum is gone. On macro liquidity, net liquidity is $5.79T, essentially flat at -0.1% week over week, with the 10-year Treasury at 4.64%. The liquidity environment leans cautious.

Where is the money going? The three strongest inflows are uranium ETFs (URA, strength 68), gold miners (GDX, 58) and gold (GLD, 57). The three heaviest outflows are FTSE China (FXI, -62), low volatility (SPLV, -57) and short-term volatility (VIXY, -52). Among broad style buckets, outflows concentrate in the Nasdaq 100 (QQQ, 12% share), and the sector rotation pairing shows money moving from utilities (XLU) into consumer discretionary (XLY), with a strength of 4.09. In trader's terms: broad indices are being trimmed slightly, and defensive low-volatility and volatility products are being cut just as hard. Money is neither chasing growth nor buying insurance — it's buying hard assets, with gold and uranium both in the top three inflows. That's a defensive-but-not-panicked positioning structure.

At the sub-sector level, there is one thread relevant to this stock. The technology sector (XLK) rotation has reached a mature stage, within which cloud computing (CLOU) on 2026-08-07 and optical modules (LITE) on 2026-08-17 moved into the strongest tier, against a backdrop of money migrating into the compute-infrastructure sub-sectors (sources: CNBC, MarketWatch, Yahoo Finance). Health care (XLV), meanwhile, has only just entered the early phase of its rotation, with six biotech-related sub-sectors moving up a tier collectively between August 18 and 21. In short, money is still buying the "AI infrastructure" theme — it has just moved from technology broadly to the specific infrastructure links in the chain.

The problem is that this stock belongs to no sector basket and appears in none of the current rotation threads. It gets no sector beta underneath it; its moves are driven almost entirely by its own news. And on August 26, the news was already cashed in once: the stock traded as much as 10.6% above the prior close intraday before finishing down 3.5% (Yahoo Finance, Robinhood, 2026-08-26). A long upper shadow running from +10.6% to -3.5% means every buyer on the announcement was absorbed by sellers — the price had run ahead of the money. The close fell back below the 50-day moving average, while the 200-day sits roughly 26% below the current price. On a weekly view, this is a spot where the news has just been digested, there is no sector support, and the short-term average has flipped from support to resistance.

Long-term (quarters to years): worth following. The reason isn't the 39% revenue growth — it's three things being true at once. Signed contracts cover the entire current enterprise value ($1.7B against $1.70B). Payback on new capacity is under two years. And the 2027 $750M/$300M target includes zero contribution from the 20,000 beds in the pipeline. The costs to accept are equally clear: GAAP net income stays ugly through these five quarters, net debt swells from $34M today toward the hundreds of millions, the company's own guardrail is net leverage well below 3x on exiting 2027, and a 15.5x current-year EV/EBITDA already treats smooth execution as the base case. Position sizing should be driven by verification points: as long as WHS utilized beds climb quarter by quarter toward 9,000, the second-half margin jump arrives on schedule, and the pipeline bed count doesn't shrink, the thesis holds. Whichever of those breaks first is the position you trim — rather than waiting for the share price to tell you.

Short-term (days to weeks): don't chase. That long upper shadow on August 26 digested the new-contract news in a single session, the close fell back below the 50-day moving average, and this stock gets no support from any sector rotation while broad indices sit neutral and money flows toward hard assets like gold and uranium. In that structure, a small-cap growth name with no basket membership lacks the follow-through buying a rebound needs. There are two things worth waiting for: the price reclaiming and holding its short-term average, or the next quarterly report clearing the second-half math hurdle. The second half of this note sets out the specific tests for both, along with three price levels, each with its own public anchor.

Two entry conditions. Neither is sufficient on its own — you need at least one money-flow condition alongside one fundamental condition.

Condition one (money flow, trackable in any daily review): SPY reclaims its 20-day moving average (766.64 as of 2026-08-26, with SPY at 766.01 that day), while the compute-infrastructure sub-sectors hold the strongest tier (cloud computing moved up on 2026-08-07, optical modules on 2026-08-17); against that backdrop, TH closes above its 50-day moving average of $17.51 for three consecutive sessions. All three have to hold simultaneously to count as a trigger — TH rebounding alone while the broad index stays below its average is a news bounce, not an entry.

Condition two (fundamental, pinned to specific numbers in the next quarterly report): Q3 total revenue of at least $120M and Adjusted EBITDA of at least $33M. Those thresholds come straight from the arithmetic in section 1 — the full-year midpoints of $440M and $110M, less the actual first-half figures of $158.2M and $28.2M, leave $281.8M and $81.8M for the second half. Distribute that across a ramp where Q3 carries roughly 40% and Q4 roughly 60%, and Q3 lands near $120M and $33M. You also want to see WHS average utilized beds rise from Q2's 4,000 to above 6,000. If Q3 revenue comes in below $110M, the full-year guidance would need Q4 to deliver more than $170M — beyond what the contract delivery schedule can reasonably support.

Three reference price levels, each tied to a publicly verifiable anchor.

Three things to check daily. First, TH's position relative to $17.51 and $14.00 — closing above $17.51 and holding for three days is the price component of condition one; closing below $14.00 triggers the early warning. Second, SPY relative to its 20-day moving average of 766.64 (2026-08-26) — reclaiming it means money flows are improving, while sustained trading below it means discounting any individual-stock rebound. Third, the 10-year Treasury yield at 4.64% (2026-08-26) — further increases push up ABL financing costs directly, which for a company borrowing hundreds of millions to build assets this year is real margin erosion.

Three things to check weekly. Whether the company files any new 8-K disclosing a contract signing — the only hard evidence of how fast that 20,000-bed pipeline converts. Whether the compute-infrastructure sub-sectors (cloud computing, optical modules) hold the strongest tier. And whether gold and uranium start ceding their top-three inflow spots to growth sub-sectors — safe-haven assets receding usually precedes money rotating back into small-cap growth.

Six numbers to check quarterly. Whether Q3 total revenue reaches $120M and Adjusted EBITDA reaches $33M — the test for condition two. Whether WHS average utilized beds rise from 4,000 to above 6,000. The actual pace of capex and the size of the jump in the ABL balance ($40M at June 30) — the most direct read on whether the leverage guardrail still holds. Whether HFS–South occupancy of 70% and the average daily rate of $72.36 keep sliding — a legacy business collapsing too fast would offset the new business's gains. Whether the balance sheet's deferred revenue and customer deposits ($21.3M at June 30) rise in step with new contracts — the evidence that customer prepayments are still coming. And how fast net leverage of 0.6x changes.

We will publish an update if any of the following occur: the definition of a newly signed contract's value weakening materially, such as shifting from committed minimum to expected revenue; another revision to full-year or 2027 guidance; the ABL balance jumping more than $200M in a single quarter; Q3 numbers falling below any of the thresholds above; the share price breaking $14.00; or either hyperscaler customer being publicly confirmed alongside an adjustment to its own capex plans.

Catalyst calendar:

We watch every line on that tracking checklist every day — whether the price breaks below April's institutional cost basis, whether SPY's relationship with its short-term average reverses, or whether the company files another contract announcement. Any trigger and we publish an update, sent out with that day's daily review, so readers don't have to sit refreshing the data themselves. Every number in this note can be verified on the corresponding data pages on the site — quarterly financials and segment readings come from the company's 8-K filings and press releases, while money flows, rotation states and macro liquidity come from that day's market data, each labeled with its own data date. Verification on something like a contract cycle happens on a quarterly rhythm; most of the price movement in between is noise. The signals that should actually change your view will usually appear in the bed count and the balance sheet first, and in the share price only afterwards.

This article was produced on the NextPick research workbench; every money-flow, tier and valuation figure it cites can be verified on the corresponding data surfaces of the site, and the moment any tracking condition listed in the report triggers, an addendum will go out together with that day's daily review.