Helios Technologies
Small valves and couplings that steer the oil moving a digger's arm, the screens and controllers that run such machines, and now couplings for water-cooled AI racks.
The scores
Growth3/4 ↑
Not 4: the record of growth is four quarters old, follows three years of decline, and 2026 sales only just regain the 2022 level.
Business quality3/4 →
Not 4: the switching cost has only the company as witness, its largest named rival competes on price, and margins fell through the downturn.
Financial position3/4 ↑
Not 4: real debt, not net cash. Covered easily - leverage below its own target range, nothing due inside the horizon, plan funded from operations.
Execution record3/4 ↑
Not 4: the record under this leadership is short, the first data centre order has already slipped once, and the last acquisition was written off.
Per share capture3/4 →
Not 4: the share count falls and buybacks beat issuance, but the anchor wants acquisitions at sensible prices and the last went to zero.
Potential2/4 →
Not 3: the data centre upside is expected, not committed - no order, volume or term - and half the 2030 plan needs unfound deals.
Risk profile3/4 →
Not 4: no customer reaches 10% of sales, but trade policy moves costs across plants in eight countries and one Electronics customer is unsized.
Valuation2/4 ↑
Not 3: at $22 for each $1 of operating profit after amortisation the price needs the 2030 plan, not just the guided year, to work.
Evidence quality3/4 →
Not 4: the financial spine is filed and cross-checked, but order intake, backlog length and the data centre pipeline have only management as a source.
If it works 1.3× · if it fails you keep about 100%.
The Verdict
We rate Helios neutral. It makes hydraulic valves and couplings for machines like diggers, plus the controllers that run them, and it has repaired itself: debt is down and margins have widened, so the business is solid. Much of that recovery is already done, and the raised forecast still implies a weaker second half. The price is full, not cheap: it counts on higher returns and a data-centre order, neither yet seen, with no cushion if neither comes. As a bet it is defensive: couplings for AI cooling might add about a third if they win, and almost nothing is lost if they fail, since the rest covers the price.
The long caseAs AI racks switch from air to water cooling, every tray needs sealed couplings; Helios can win a share if buyers approve it beside makers already shipping.
The evidence under this - 8 claims
- Management has beaten its own quarterly guidance in seven of the last eight prints, so the guided second half may be conservative in the same way the first half was.
- The comparison flatters 2025 slightly: the second half of 2025 still contained Custom Fluidpower earnings for part of Q3, and that business has since been sold.
- Management said on the same call that it 'would be disappointed' if no order came through in the second half, and said the same thing on the prior quarter's call - the date has moved once already.
- The sale is to integrators building cooling racks, but the product must first be validated by a hyperscaler, and management confirmed the samples out are not primarily with hyperscalers.
- Stäubli's competing couplings have been on NVIDIA's recommended vendor list since at least November 2025.
- The research pack's own multiples are struck on an enterprise value that treats the company as holding net cash, so every enterprise-value multiple there reads about 12% cheaper than the filed balance sheet supports - and its historical medians are computed the same way, so the comparison against the company's own history may still be sound even though the level is not.
- The trailing reported figure is depressed by the $25.9M goodwill write-off in Q3 2025, which will drop out of the trailing window after the November print.
The company
A digger's arm moves because oil under pressure is steered through valves. Helios makes small valves that screw into a machined block, and couplings that let a hose be unplugged under pressure without spilling. That is 63.1% of sales. The rest is electronics: rugged displays, engine controllers and wiring for the same machines, and control systems for hot tubs.
Its buyers are equipment makers and the distributors who stock their parts. The alternative is a whole hydraulic system from a giant such as Parker Hannifin or Bosch Rexroth.
The evidence under this - 3 claims
How it makes money
Helios is paid per part, and the part is chosen when a machine is designed, then bought for every machine built. No customer reached 10% of 2025 sales. Hydraulics carries the profit: 19.7 cents of each sales dollar in the June quarter, against 13.1 cents in electronics.
The evidence under this - 3 claims
- Parker-Hannifin, the largest company Helios names as a competitor, describes competition in the same market as turning on 'product quality and innovation, customer experience, manufacturing and distribution capability, aftermarket support, and price competitiveness' - naming price and never naming switching costs.
- Helios's own 10-K names low-cost producers, Winner and Valvole Italia, that 'typically attempt to copy our products', which is what a copyable product looks like.
- Group operating margin fell to 7.9% of sales in fiscal 2025 and Hydraulics went with it, which is not how a business whose customers cannot leave behaves through a downturn.
Where the money comes from, stream by stream
| Stream | Share of sales (fiscal 2025) | How it earns it |
|---|---|---|
| Hydraulics | 64.5% | Screw-in cartridge valves and manifolds and Faster quick-release couplings, sold through distributors and to machine makers for hydraulic circuits. |
| Electronics | 35.5% | Displays, controllers, gauges and harnesses for engine-driven machines, boats and specialty vehicles, and Balboa spa control systems, mostly sold to equipment makers. |
Who runs it
Sean Bagan, who joined as finance chief from Polaris in 2023, has run Helios since the board fired his predecessor in mid-2024. Since then debt has fallen and the business that did not fit was sold. The finance job changed hands twice in late 2025. Insiders own 0.5% of the shares. The bonus rides on profit, sales and cash; the long-term awards on three-year sales and earnings. Nothing pays on return on capital, and the last three-year awards paid nothing.
The evidence under this - 4 claims
The one sizeable deal of the window lost its whole $25.9M of goodwill within 28 months (C13), and the 2030 plan needs acquisitions again (C14).
Why it wins, and who could beat it
Why customers come back
Once a valve's cavity is machined into a block, or a controller is programmed for the buyer's machine, changing supplier means re-machining, retesting and rewriting software. Helios says that cost is what holds its position.
Its biggest rival reads the market differently: Parker Hannifin says competition turns on quality, service and price, and never mentions the cost of switching. The only outside evidence is the margin, and gross margin has widened four quarters running, to 34.6%.
The evidence under this - 3 claims
- Parker-Hannifin, the largest company Helios names as a competitor, describes competition in the same market as turning on 'product quality and innovation, customer experience, manufacturing and distribution capability, aftermarket support, and price competitiveness' - naming price and never naming switching costs.
- Helios's own 10-K names low-cost producers, Winner and Valvole Italia, that 'typically attempt to copy our products', which is what a copyable product looks like.
- Group operating margin fell to 7.9% of sales in fiscal 2025 and Hydraulics went with it, which is not how a business whose customers cannot leave behaves through a downturn.
The threat
In data centres, the coupling Helios hopes to sell follows an open standard: its dimensions are published so any maker's part fits any other's, and Stäubli already sells one NVIDIA has approved. In machines, the slower threats are electric drives such as Bosch Rexroth's, which replace oil circuits with motors, and low-cost copiers of the same valves.
The evidence under this - 4 claims
The rivals: PH, APH, and 3 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| Private | Stäubli UQD and UQDB couplingsPriced against it: Private | Data centre liquid cooling couplings | On sale | Private |
| PH · APH | CEJN, Parker Hannifin and Amphenol universal quick disconnectsPriced against it: Same product | Data centre liquid cooling couplings | On sale | Same product |
| Private | Bosch Rexroth eLION electric drivesPriced against it: Private | Hydraulics and Electronics, without data centre couplings | On sale | Private |
| Private | Winner and Valvole Italia copied cartridge valvesPriced against it: Private | Hydraulics and Electronics, without data centre couplings | On sale | Private |
How big it could get
How big this could get
Dell'Oro, a research firm, puts spending on data-centre liquid cooling close to $3B in 2025, roughly double the year before, and near $7B by 2029. Couplings are one part of that, and each rack needs many: an NVIDIA GB300-class rack plugs water into 27 trays. But nobody publishes how many such racks ship or what a coupling costs, so the coupling market is smaller than $3B by an unknown amount.
The evidence under this - 2 claims
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Hydraulics | 63% | +3.9% | +6.8% | +8.3% |
| Electronics | 37% | +19.4% | +24.2% | +25.1% |
| Whole company | 100% | +9.1% | +12.8% | +14.0% |
The filed split through 2026-07-04. These rates are not carried forward: the reading below says why.
How we read it: The blend. Electronics' 24% is a rebound, not a shift: $334M in 2022, $269M in 2024, $332M trailing, with marine still depressed. The couplings the case rides sit in Hydraulics, slowed by the CFP sale.
The results
The numbers
Sales grew 14% over the twelve months to Jul 2026, faster than the 4.1% in the year to Jan 2026.
Operating margin is 10%, about where it was (9.6%) in the year to Dec 2023.
Running costs take 24¢ of each sales dollar, up from 19¢ in the year to Jan 2022.
Each dollar put into the business earns 5.9% a year after tax, down from 14% in the year to Jan 2022.
It turned 12% of sales into free cash, $107M over the last twelve months.
It owes $158M more than the cash it holds, about 1.2 years of the cash its operations bring in.
The share count rose 2.2% in 4.5 years, and pay in shares costs 0.7% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-04. Fiscal years end in January.
| Jan 2022 | Dec 2022 | Dec 2023 | Dec 2024 | Jan 2026 | 12 mo to Jul 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $869M | $885M | $836M | $806M | $839M | $891M |
| Sales growth | +66% | +1.9% | −5.6% | −3.6% | +4.1% | +14% |
| Profit | ||||||
| Gross margin | 36% | 34% | 31% | 31% | 32% | 34% |
| Operating margin | 17% | 16% | 9.6% | 10% | 7.9% | 10% |
| Net margin | 12% | 11% | 4.5% | 4.8% | 5.8% | 8.0% |
| What growth costs, per sales dollar | ||||||
| Research and development | 1.9% | 2.0% | 2.3% | 2.5% | 2.3% | - |
| Selling, general and admin | 15% | 15% | 18% | 17% | 18% | 17% |
| Pay in shares | 1.0% | 1.0% | 1.4% | 0.4% | 0.6% | 0.7% |
| Capital spending | 3.1% | 3.6% | 4.1% | 3.4% | 2.8% | 3.4% |
| Cash | ||||||
| Cash from operations | $113M | $110M | $83.9M | $122M | $127M | $137M |
| Free cash flow | $86.3M | $78.0M | $49.6M | $95.1M | $104M | $107M |
| Balance sheet | ||||||
| Cash and short-term investments | $28.5M | $43.7M | $32.4M | $44.1M | $73.0M | $68.0M |
| Debt | $202M | $183M | $322M | $299M | $262M | $226M |
| Net cash | −$174M | −$140M | −$289M | −$255M | −$189M | −$158M |
| Returns and shares | ||||||
| Return on capital | 14% | 12% | 5.7% | 5.4% | 4.3% | 5.9% |
| Shares (diluted) | 32.5M | 32.6M | 33.0M | 33.3M | 33.3M | 33.2M |
What growth costs
Growth is making Helios more profitable mainly in its factories: more volume spreads the fixed cost of the plants. Overhead has grown about as fast as sales, and still takes more of each sales dollar than before the downturn. The 2023 slump showed the same lever works in reverse.
The evidence under this - 2 claims
- At trailing sales of $891.4M, about fiscal 2022's $885.4M, overhead still takes about two cents more of each sales dollar than it did then, so the recovery has not yet bought back the overhead added in the downturn.
- Part of the gross-margin gain is net tariff refunds, about $1M in Q2 2026 (C21), which do not recur by design.
The last report (10 Aug)
Quarter ended July 2026, filed figures.
The second quarter of 2026 was a good one. Gross margin reached 34.6%, orders grew by double digits for the fourth quarter running, and the full-year forecast rose again, to $880M to $900M of sales.
But the forecast holds a catch. It implies second-half adjusted earnings of about $1.47 a share, against $1.53 a year earlier: the recovery flattening.
What it did to the thesisIt confirmed the recovery and the stronger balance sheet are real. It also showed the earnings rebound flattening, since the raised forecast implies a weaker second half than last year's. The data centre order the growth story needs is still not signed, so the view stays Neutral.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.88 against $0.80 expected; $0.66 as filed, the gap being what the company adds back.
The evidence under this - 4 claims
- Order intake is not a filed figure and appears in neither the 10-Q nor the results release, so there is no way to check the rate or the base it is struck on.
- The comparisons are against a period in which the business was still shrinking.
- The same release guides Q3 2026 adjusted earnings to $0.70-0.77, only 2% above the year before at the midpoint, and an adjusted cash-profit margin 30 basis points lower than a year earlier - so the raise is a first-half event.
- Management's own reason for the earlier caution still stands: order visibility is a little over one quarter, so the second half is guided rather than booked.
- Management has beaten its own quarterly guidance in seven of the last eight prints, so the guided second half may be conservative in the same way the first half was.
- The comparison flatters 2025 slightly: the second half of 2025 still contained Custom Fluidpower earnings for part of Q3, and that business has since been sold.
What management did with the money
It has used the recovery to cut debt: net debt of $263.5M is 1.4 times a year's cash profit, against 2.6 times a year earlier. It sold a business and raised the dividend 33%. It also wrote off one acquisition's entire $25.9M of goodwill, 28 months after buying it.
The evidence under this - 4 claims
The price, and where we disagree
In short. Full. Fair for the recovery already delivered, dear for the one still to come: returns on money invested must rise and a data-centre order must arrive, and neither has happened yet.
The shares have fallen 24.7% from their high of the past year, 14.1% in the last month alone, though they are still up 31.4% on a year ago.
The next answer comes on 2 November with the September-quarter results. Adjusted earnings of $0.70 to $0.77 a share are forecast; the bigger question is whether a data-centre order has arrived. One with a volume attached would be the first hard evidence for the new market.
The evidence under this - 3 claims
- Management said on the call that the second-half comparisons are the hard ones and that four of the last five years put 52% to 54% of revenue in the first half, so the shape is seasonal rather than a deterioration.
- July was described on the call as the company's best month ever for both revenue and order intake, which sits awkwardly with a quarter guided to 2% earnings growth.
- Management said on the same call that it 'would be disappointed' if no order came through in the second half, and said the same thing on the prior quarter's call - the date has moved once already.
- The sale is to integrators building cooling racks, but the product must first be validated by a hyperscaler, and management confirmed the samples out are not primarily with hyperscalers.
- Stäubli's competing couplings have been on NVIDIA's recommended vendor list since at least November 2025.
How we read the multiple
Counting debt, Helios costs about $2.63B: about $33 for each dollar of last year's profit and about $14 for each dollar of cash profit, both near its usual levels. On this year's expected operating profit after the cost of past acquisitions, it is about $22.
That buys a business earning about 7.7 cents a year on each dollar invested in it. Its 2030 plan aims for a return in the teens on $1.6B of sales, roughly double today's, partly from acquisitions not yet named.
Against its own history. Close to its own normal for each dollar of cash profit, and cheaper on earnings. The price is not stretched against its past; the question is whether returns rise.
The evidence under this - 3 claims
- The research pack's own multiples are struck on an enterprise value that treats the company as holding net cash, so every enterprise-value multiple there reads about 12% cheaper than the filed balance sheet supports - and its historical medians are computed the same way, so the comparison against the company's own history may still be sound even though the level is not.
- The trailing reported figure is depressed by the $25.9M goodwill write-off in Q3 2025, which will drop out of the trailing window after the November print.
- The denominator is dominated by acquired intangibles: on tangible capital the operating businesses earn a far higher return, and it is the price paid rather than the operations that depresses the ratio.
- 2026 is the first year of recovery from a three-year downturn, so this is not a mid-cycle return; the research pack's own trailing figure of 5.1% is lower still and struck on a worse year.
- Helios has completed no acquisition since May 2023, and its chief executive said on the Q2 2026 call that neither he nor the chief financial officer has done one at Helios.
- The one recent deal that can be scored had its goodwill written to zero within 28 months.
- The return target is roughly double what the current guided year implies, and the 10-K says the acquisition approach 'has evolved from lessons we have learned ... from our past acquisitions'.
Priced against its rivals
Only Parker-Hannifin sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
Also checked and set aside: Eaton (a substitute, not the same product); Garmin (a substitute, not the same product).
Private or foreign rivals that cannot be priced: Danfoss A/S, Bosch Rexroth AG, HydraForce, Delta Power Company, Stucchi S.p.A., CEJN AB, Winner, Valvole Italia, Continental AG, Robert Bosch GmbH.
The bet: what it pays if it works
This bet is on one business only: couplings that carry cooling water to chips in artificial-intelligence racks. Everything else Helios makes is counted separately, at the price the market pays for it today, and on its own it covers the whole price.
So the couplings come free, but even a good outcome adds little: winning a small slice of a fast-growing market leaves the company worth only a little more than it costs, which makes this an expensive bet. A rival is already approved, and Helios has no order yet.
| Data centre liquid cooling couplings revenue in two to three years | Points of the 2029 data centre liquid cooling equipment market needed | At the stated multiple |
|---|---|---|
| $70.00M (the big win) | about 1 at 100% kept | 1.3× |
| $100.00M (market forecast) | about 1 at 100% kept | 1.3× |
| $140.00M | about 2 at 100% kept | 1.4× |
About 100,000 NVL72-class racks a year by 2029 at roughly 250 coupling pairs and $50 a pair is a $1.25B coupling market; 8%, below an equal split among ten-plus makers, is $100 million. On that forecast, $100.00M pays 1.3× from today's value: an Expensive bet on the market case. Data centre liquid cooling couplings would be worth about $300.00M, and the price pays nothing for it.
To pay 2× (a Fair bet) it would need about $581.01M of Data centre liquid cooling couplings revenue (about 8 points of the 2029 data centre liquid cooling equipment market); to pay 5× (a Cheap bet), about $2.77B of Data centre liquid cooling couplings revenue (about 40 points of the 2029 data centre liquid cooling equipment market), at the same multiple.
Each row is Data centre liquid cooling couplings alone, plus the rest of the company (Hydraulics and Electronics, without data centre couplings) counted at about $2.64B.
The steps it has to take
- Done Four coupling ranges launched to the Open Compute standard: UQD, UQDB, LQC and FFC
- Done North American Faster production consolidated into Toledo, Ohio
- Done Samples placed with about a dozen prospective customers
- Slipped A first data centre coupling order announced due 2026-11-02
- Not started Helios couplings approved by NVIDIA, as Stäubli's already are due 2027-06
- Not started An order disclosed with a stated volume and term due 2027-06
- Not started Coupling sales reported for 2027, as management expects due 2028-03
- Not started Coupling revenue near one point of the 2029 market, about $70M a year due 2029-03
The detail behind the label
The evidence under this - 4 claims
- Management said on the same call that it 'would be disappointed' if no order came through in the second half, and said the same thing on the prior quarter's call - the date has moved once already.
- The sale is to integrators building cooling racks, but the product must first be validated by a hyperscaler, and management confirmed the samples out are not primarily with hyperscalers.
- Stäubli's competing couplings have been on NVIDIA's recommended vendor list since at least November 2025.
- The research pack's own multiples are struck on an enterprise value that treats the company as holding net cash, so every enterprise-value multiple there reads about 12% cheaper than the filed balance sheet supports - and its historical medians are computed the same way, so the comparison against the company's own history may still be sound even though the level is not.
- The trailing reported figure is depressed by the $25.9M goodwill write-off in Q3 2025, which will drop out of the trailing window after the November print.
Where we disagree with the market
A recovering cyclical with a free option on AI data-centre cooling, after two forecast raises and four quarters of wider margins.
The debt repair is the lasting gain. The earnings recovery is mostly done, and the data-centre option is weaker than its billing.
Two public facts rarely get backed out. The raised forecast still implies a weaker second half than last year's. And the data-centre coupling is built to a published standard, so the buyer can plug in anyone's part.
We would be wrong if a data-centre order arrives with a stated volume and term, or if second-half adjusted earnings beat last year's $1.53 a share.
What analysts say, and the evidence
The evidence under this - 4 claims
- Management has beaten its own quarterly guidance in seven of the last eight prints, so the guided second half may be conservative in the same way the first half was.
- The comparison flatters 2025 slightly: the second half of 2025 still contained Custom Fluidpower earnings for part of Q3, and that business has since been sold.
- Management said on the same call that it 'would be disappointed' if no order came through in the second half, and said the same thing on the prior quarter's call - the date has moved once already.
- The sale is to integrators building cooling racks, but the product must first be validated by a hyperscaler, and management confirmed the samples out are not primarily with hyperscalers.
- Stäubli's competing couplings have been on NVIDIA's recommended vendor list since at least November 2025.
What could go wrong, and how we would know
The risks
- No order yet. Helios has samples with about a dozen prospective data-centre customers, no such sales in this year's forecast, and nothing signed.
- Acquisitions that sour. The 2030 target counts on purchases not yet identified, and the last acquisition's goodwill was written off in full.
- Little behind the accounts. Acquired intangibles are $841.8M of $1.52B of assets; only about $106.5M of shareholders' equity rests on anything else.
- A cyclical customer base. Machine makers cut orders in downturns, and boating, a large electronics market, shows no turn yet.
The evidence under this - 5 claims
- Management said on the same call that it 'would be disappointed' if no order came through in the second half, and said the same thing on the prior quarter's call - the date has moved once already.
- The sale is to integrators building cooling racks, but the product must first be validated by a hyperscaler, and management confirmed the samples out are not primarily with hyperscalers.
- Stäubli's competing couplings have been on NVIDIA's recommended vendor list since at least November 2025.
- Helios has completed no acquisition since May 2023, and its chief executive said on the Q2 2026 call that neither he nor the chief financial officer has done one at Helios.
- The one recent deal that can be scored had its goodwill written to zero within 28 months.
- The return target is roughly double what the current guided year implies, and the 10-K says the acquisition approach 'has evolved from lessons we have learned ... from our past acquisitions'.
- Electronics still grew 19% year over year in Q2 2026 with a record second quarter at Enovation Controls, so the segment is outgrowing its weakest market.
- Management attributes the weakness to interest rates on financed equipment, which is a cyclical cause rather than a structural one.
What we still do not know
- How many liquid-cooled racks ship each year, and what a coupling sells for. Neither is published, so the data-centre prize cannot be sized.
- Whether customers really are locked in. Helios discloses no count of design wins, programme lengths or repeat orders.
- Which acquisitions would fill the gap to the 2030 target.
The evidence under this - 4 claims
- Parker-Hannifin, the largest company Helios names as a competitor, describes competition in the same market as turning on 'product quality and innovation, customer experience, manufacturing and distribution capability, aftermarket support, and price competitiveness' - naming price and never naming switching costs.
- Helios's own 10-K names low-cost producers, Winner and Valvole Italia, that 'typically attempt to copy our products', which is what a copyable product looks like.
- Group operating margin fell to 7.9% of sales in fiscal 2025 and Hydraulics went with it, which is not how a business whose customers cannot leave behaves through a downturn.
- Helios has completed no acquisition since May 2023, and its chief executive said on the Q2 2026 call that neither he nor the chief financial officer has done one at Helios.
- The one recent deal that can be scored had its goodwill written to zero within 28 months.
- The return target is roughly double what the current guided year implies, and the 10-K says the acquisition approach 'has evolved from lessons we have learned ... from our past acquisitions'.
What would change our view
- Neutral→AttractiveReturns rise and the order is filed
It needs all of:
- ○A data-centre coupling order with a stated volume and term is disclosed by June 2027, not just described.checked at the next results, due by 2 Nov
- ○The return on invested money rises from about 7.7% toward the teens, with hydraulics still earning above a 15% margin.checked at the next results, due by 2 Nov
- Neutral→UnattractiveThe guidance or the order book slips
It needs any of:
- ○Second-half 2026 adjusted earnings come in below $1.37 a share, the bottom of the forecast.checked at the next results, due by 2 Nov
- ○Orders fall from a year earlier for two quarters running, or an acquisition is written down within three years.checked at the next results, due by 2 Nov
We would also be wrong if
- No data-centre order with a stated volume and term by June 2027.
- Hydraulics margin falls back below 15% for two quarters running.
What is coming
- 2 November 2026 Q3 2026 results: Adjusted EPS against the guided $0.70 to $0.77, whether the second half softens as guided, and whether a data-centre coupling order is named with a volume.
Every dated event ahead
- 2 November 2026 Q3 2026 results: Adjusted EPS against the guided $0.70 to $0.77, whether the second half softens as guided, and whether a data-centre coupling order is named with a volume.
- March 2027 Fiscal 2026 results, 10-K and 2027 guidance: Whether the year lands inside the raised $880M to $900M range and whether 2027 guidance carries any data-centre coupling sales or an acquisition toward the 2030 plan.
- December 2027 First data-centre coupling sales: Whether samples with about a dozen prospects become revenue, and at what scale against the market case of about $100M of coupling sales in 2029.
Since we wrote this
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 24 SepNewsAn immaterial asset purchase and an unsized motorcycle contract; case unchanged. View heldHelios bought some assets of Radio Sound, an audio-electronics maker, alongside a multi-year Indian Motorcycle contract.
- 6 SepThesis writtenNeutral, medium convictionThe recovery and the debt repair are real and mostly delivered, but the second half is guided flat to lower and the data-centre option enters an open-standard market.
Sources 26
Sources - 26
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-03-03the business and segment descriptions, the named competitor list, the switching-cost assertion, the customer-concentration statement, the acquisition and divestiture history, and the standardisation risk phrasebacksC01C10C13C22C23C31C37C40C50C51C52C55C56
- S02primarycompany filing2026-08-11the balance sheet including both debt lines, the income statement, the cash flow statement's depreciation and amortisation, the share counts and the geographic splitbacksC02C03C11C12C21C23C24C25C50C52C56
- S03primarycompany filing2026-08-10the GAAP-to-adjusted earnings reconciliation line by line, the raised full-year and third-quarter guidance, the segment commentary and the leverage and buyback figuresbacksC04C06C07C09C11C16C21C24C25C35C39
- S04primarytranscript2026-08-11Helios Q2 2026 earnings call transcript, held 2026-08-11the data centre pipeline, the backlog length, the capacity and capital-spending commentary and the acquisition philosophy from management; and, as a separate origin, the questions from four analysts pressing on second-half cadence, the data centre addressable market and capital allocation
This represents our single largest opportunity across Helios today ... we now have samples out with about a dozen prospective customers that are sampling our products
backsC09C16C19C20C32C35C39 - S05primarycompany filing2026-03-02the fiscal 2025 reported and adjusted earnings, the pro forma sales base excluding the divested business, and the original 2026 guidance the August raise is measured againstbacksC05C07C22
- S06primarycompany filing2025-11-04the $25.9M goodwill impairment, which reporting unit it belonged to and why, and the $18.8M gain on the Custom Fluidpower sale sitting below the operating linebacksC13C22
- S07secondaryvendor data2026-09-06Research pack, frozen at content/HLIO/research-pack.jsonprice, market value, returns, the volatility bucket, the analyst split, the Form 4 transcription, short interest and the filer's own XBRL segment splitbacksC02C03C27C28C29
- S08primarycompany release2026-03-20the 2030 financial targets and the 33% dividend increase
Sales of $1.6 billion (including acquisitions), Adjusted EBITDA margin of 25%+ of sales, Adjusted Operating Income margin of 20% of sales, Return on Invested Capital in the low- to mid-teens
backsC14C25 - S09primarycompany release2026-05-05the four launched coupling ranges and the standard they are built to, and the absence of any named customer, volume or term
compliant with OCP (Open Compute Project) standards for liquid cooling in data centers
backsC17C38 - S10primaryindustry data2026-09-06CEJN - Universal Quick Disconnect product documentation CEJNa competitor's own statement that the interface Helios is entering on is intermateable across suppliers
As an open standard, any manufacturer's products are connectable by following the specifications of the UQD standard.
backsC17 - S11primarycompany release2025-11-12that a competitor already holds the qualification Helios is still seeking, and has clean-room capacity in the United States for it
remain approved by NVIDIA and are included in NVIDIA's Recommended Vendor List (RVL)
backsC18 - S12primaryindustry data2026-09-06the tray count per rack and the statement that the midplane routes water to each tray through quick connects
18 GPU Compute trays, 9 Switch trays and either 6 or 8 Power Shelves all connected via a rear cable cartridge backplane ... The midplane in the chassis routes the water via QuickConnects to each of the Compute Tray, and each of the NVL Switch Trays
backsC34 - S13primaryindustry data2026-01-08Dell'Oro Group - data centre liquid cooling market forecast Dell'Oro Groupthe only outside sizing of the market the data centre couplings sell into, and the absence of a component breakdown within it
The Liquid Cooling market is expected to roughly double in 2025, reaching close to $3 billion in revenue, before continuing to scale toward approximately $7 billion by 2029
backsC33 - S14primaryindustry data2026-09-06the attacker's own framing of electrification as sold alongside hydraulics rather than as replacing it
Whether driving or working, hydraulic, hybrid or fully electric
backsC30 - S15primarycompany filing2026-02-26that Eaton no longer describes an industrial or mobile hydraulics business, which makes one name on Helios's own competitor list stalebacksC36
- S16primarycompany filing2026-08-21the confirmation that Parker sells hydraulic valves to the same buyers, the 67/33 split between its industrial and aerospace segments, and its own account of what competition in this market turns onbacksC41
- S17primarycompany filing2026-05-11the middle step of the guidance path - sales raised to $840-870M and adjusted earnings to $2.75-3.00 in May - plus net debt of $284M and leverage of 1.6x at the first quarterbacksC07
- S18secondaryanalyst2026-07-12BigGo Finance - GB Racks to Ship 80,000 Units Annually! Morgan Stanley Names Wistron Top AI Manufacturing Pick (report of a Morgan Stanley note) Morgan Stanley, reported by BigGo Financea bank's count of NVIDIA GB200/GB300 NVL72 rack shipments, about 29,000 in 2025 and 70,000 to 80,000 forecast for 2026, with ODM-by-ODM monthly output; the note itself is paid, so this is a report of it, read through ir_fetch
The firm estimates that full-year shipments of GB200 and GB300 NVL72 racks will reach 70,000 to 80,000 units this year, representing a multiple-fold increase from approximately 29,000 units last year.
backsC42 - S19primaryindustry data2026-08-17an independent research firm's liquid-cooling penetration among AI chips (about 33% in 2025, 53% in 2026, near 60% in 2027) and NVIDIA rack-scale shipments doubling in 2026 and growing more than 30% in 2027; read through ir_fetch
TrendForce notes that accelerated AI investment by Tier 2 data center operators and growing demand from Chinese CSPs for overseas AI projects are expected to double shipments of NVIDIA’s GB/VR rack-scale solutions in 2026. Despite the possibility of delays to Kyber NVL144 in 2027, strong AI infrastructure investment is expected to sustain overall GPU rack demand, with shipments still projected to grow by more than 30% YoY.
backsC43 - S20secondaryother2026-07Netonx Europe - How Many UQD Couplings per NVL72 Rack? GB200 vs GB300 (Per-Tray Math) Netonx Europe (Green MeOH GmbH), a UQD distributorthe coupling count per NVL72 rack, about 126 pairs for GB200 and 250 for GB300, built tray by tray and labelled industry estimates by a distributor that sells a competing coupling; read through ir_fetch
In short: Per NVIDIA NVL72 rack, a GB200 build uses about 126 UQD coupling pairs and a GB300 build about 250 pairs (industry estimates) - the jump comes from more cold-plate loops per GPU tray.
backsC44 - S21secondarynews2025-09-07傅里叶的猫 (Tencent News) - GB200 GB300液冷价值量拆解 (liquid-cooling content teardown of GB200 and GB300 racks) 傅里叶的猫, a writer who says he distributes NVIDIA serversthe only per-coupling price found: $45 a pair on GB200 and $55 on GB300, the quick-disconnect counts per rack, the cold-plate, hose and manifold content beside them, and the shift of GB300 coupling supply to NVUQD from Cooler Master, AVC and Fositek; read through ir_fetch
GB200 和 GB300 的快接头需求因产品设计不同而有所差异。GB200 包含 126 个计算部分快接头和 18 个 Switch 部分快接头;GB300 的计算部分快接头数量翻倍至 252 个,Switch 部分仍为 18 个,总量明显高于 GB200。 在价格方面,GB200 的快接头价格为 45 美元/对,而 GB300 的价格略升至 55 美元/对。 GB300 的快插头方案升级为 NVUQD,供应商体系发生重大变化,欧美企业的份额大幅减少,Cooler Master、AVC、富士达等厂商成为主要供应商。由于产能限制,Cooler Master 还联合英维克参与快插头供应,以补充产能需求。
backsC45 - S22primaryindustry data2024-10-17who supplied quick disconnects for GB200 racks (CPC, Parker, Danfoss, Stäubli), that they were frequently out of stock, and that LOTES and Fositek were qualifying to join; read through ir_fetch
On the other hand, quick disconnect (QD) components are critical for preventing leakage and have frequently been reported as out of stock. Taiwanese companies like LOTES and Fositek are reportedly in the validation phase, and by the first half of 2025, they are expected to join the list of QD suppliers, to compete with major companies such as CPC, Parker Hannifin, Denfoss, and Staubli.
backsC46 - S23secondaryindustry data2025MarketsandMarkets - UQD Coupling Market, Global Forecast to 2030 (report summary page) MarketsandMarketsa paid market report's size for UQD couplings across all applications, $0.39B in 2025 to $0.75B in 2030, naming Stäubli, Danfoss, Parker, CPC and Gates as leaders; the free summary page, undated beyond its 2024 base year, read through ir_fetch
According to Marketsandmarkets, the universal quick disconnect (UQD) market size is expected to grow from USD 0.39 billion in 2025 to USD 0.75 billion by 2030 , at a CAGR of 14.3% from 2025 to 2030. The market is driven by the rapid expansion of data centers and high-performance computing, which require advanced liquid cooling solutions to prevent overheating in dense environments. Increasing adoption in energy storage, electronics cooling, and industrial automation also demands efficient, tool-free connections for minimal downtime.
backsC47 - S24primarytranscript2026-08-06Parker-Hannifin fiscal 2026 fourth-quarter earnings call transcript Parker-Hannifinthe size of a leading coupling rival's whole data-centre line, about 1.5% of fiscal 2026 sales of $21.5B, and what it spans; read through transcript_fetch (FMP)
It's about 1.5% of our sales, and we've been previously saying 1%. So it is growing, but it's not quite large enough yet to have its own market vertical. This is a great story for our interconnected technologies because you mentioned a few of our products, but it's hoses, couplings, manifolds, fittings -- engineered materials for thermal management.
backsC48 - S25primarycompany filing2025-02-25the fiscal 2022 segment table: Electronics net sales from external customers of $334.1MbacksC50C52C56
- S26primarycompany filing2026-05-04beneficial ownership as of April 20, 2026, the officers' biographies and start dates, the 2025 bonus metrics, goals and payouts, the long-term award design, and the zero payout of the 2023-2025 performance unitsbacksC53C54C55