Thermo Fisher Scientific

There is an old line about gold rushes — that the reliable money was made selling picks and shovels rather than digging. Thermo Fisher Scientific is the picks-and-shovels business of modern biology. It does not discover drugs. It sells the reagents, the instruments, the plastics, the cell-culture media and the laboratory services to everybody who does. Whichever cancer drug wins, whichever obesity drug wins, whichever biotech goes bankrupt trying — Thermo Fisher was paid along the way.

That is a genuinely attractive shape for a business, and it is why this company came up when we went looking for quality that might have fallen to a sensible price. It had a spectacular pandemic, a hard fall afterwards, and a wave of frightening headlines in 2025 about American research funding being cut. On the face of it, exactly the setup we were hunting.

So let us deal with the premise before anything else, because it does not survive contact with the data. Thermo Fisher trades today at 30.3 times its trailing accounting earnings. Its average over the past decade is roughly 31.6 times. The shares are about 16% below their December 2021 high and more than a fifth above where they stood a year ago. The moment this company was cheap was April 2025, when it traded at $446. It is not cheap now. It is ordinary.

Thermo Fisher spent about thirty billion dollars on acquisitions between 2020 and 2025. Its free cash flow at the end of that was lower than at the start. Everything worth saying about this company is contained in those two sentences.

That is not an accusation of incompetence — quite the opposite. The businesses Thermo Fisher owns are excellent. Its Life Sciences Solutions division earns a 37% operating margin, protected by switching costs that are enforced by drug regulators rather than by customer preference. That is close to a toll bridge, and we will spend Part V explaining exactly why.

The problem is arithmetic, not quality. Thermo Fisher has been buying wonderful businesses at prices fair enough that the sellers kept most of the wonder. Revenue has more than doubled in a decade. Return on capital has not followed. And an investor buying today is paying a full multiple for the aggregate rather than a discount for the parts.

Two things stand out from that table, and they point in opposite directions.

The first is that the deals which built the quality were the small ones. Life Technologies, Dionex, FEI, PeproTech, Olink, Solventum's filtration business — consumables and platforms bolted onto an installed base. That is the razor-and-blades logic working exactly as advertised, and it is why Life Sciences Solutions earns nearly forty cents of operating profit on the dollar.

The second is that the money did not go there. Patheon at $7.2 billion, PPD at $17.4 billion and Clario at $8.9 billion add to roughly $33.5 billion deployed into contract manufacturing and clinical research — businesses that sit inside a segment earning a 14.0% margin, where contracts are competitively bid and backlog can be cancelled. The serial acquirer has been buying down its own margin structure.

After three years in which the market's central complaint was the PPD acquisition, management's answer was to write a nine-billion-dollar cheque for more of the same. That is either high conviction or what Buffett calls the institutional imperative. Part VIII offers the evidence; we let you decide.

The most quoted statistic about Thermo Fisher is that roughly 84% of its revenue is recurring — consumables and services rather than one-off capital equipment. The figure is accurate. Here is the FY2025 disaggregation, which reconciles exactly to reported revenue:

But the word "recurring" is doing an enormous amount of work in that 84%, and separating the two halves is the single most useful thing an investor can do with this company.

Consumables — about 42% — are the genuine article. Reagents, cell-culture media, chromatography columns, filters, antibodies, laboratory plastics. Once a customer's process depends on them, changing supplier is expensive and slow for reasons we set out in Part V. This is annuity revenue in the proper sense.

Services — the other 42% — are not the same thing at all. That line contains PPD's clinical research, Patheon's contract manufacturing, the newly acquired Clario, and instrument service contracts. Clinical trials and manufacturing campaigns are projects. They are competitively bid, they are re-tendered, sponsors routinely use more than one provider, and backlog gets cancelled when a drug fails. Calling that "recurring" in the same breath as Gibco cell-culture media is the most common analytical error made about this company.

So the honest version reads: roughly 42% is a true razor blade, perhaps another ten to fifteen points is sticky aftermarket service and diagnostics reagents, and something over 30% is project work with ordinary switching costs. Eighty-four percent recurring is defensible and it flatters the business.

Thermo Fisher's profit pool is not where its revenue is. Life Sciences Solutions produces 38% of segment profit on 23% of revenue; Laboratory Products & Biopharma Services produces 34% of profit on 56% of revenue. And the acquisition budget has gone overwhelmingly to the second.

One more number worth holding on to: research and development ran at $1.40 billion in 2025 — about 3.1% of revenue. For a company at the centre of the life sciences, that is remarkably low. Thermo Fisher does not, on the whole, invent things. It buys the companies that invented them. That is a legitimate strategy, and it is the reason Part VIII matters more here than it would at almost any other business we have examined.

Averaging the moat across Thermo Fisher is the central analytical trap with this company. It is genuinely wide in one segment and genuinely narrow in the segment that produces most of the revenue. So we rate it segment by segment.

In a regulated environment, changing supplier is not a purchasing decision. It is a regulatory event.

When a biological drug is manufactured, the cell-culture medium, the purification resin and the single-use assemblies used to make it are written into the filing submitted to the regulator. When an analytical method is validated — the assay that proves each batch is what it claims to be — the instrument and the column are specified in that method. Changing any of it requires revalidation, comparability studies and frequently an amendment filed with the agency and reviewed by it.

The cost of switching is therefore measured in months of delay and regulatory risk to a drug that is already earning money — not in the price difference on a bottle of reagent. No procurement department saves enough on consumables to justify that. So the supplier written into the filing gets paid on every batch, for as long as the drug is sold.

That is structurally a royalty on the pharmaceutical industry's output, indifferent to which drug wins. It is exactly the kind of asset Buffett describes as a toll bridge, and it is why Life Sciences Solutions earns 37% margins while the rest of the company earns half that.

The toll bridge is real. It is also 23% of the company. The discipline this report tries to keep is refusing to let the quality of that segment launder the other 77%.

The Danaher comparison is the one worth dwelling on, because the two companies started from the same idea and have made opposite decisions with it.

Both are serial acquirers. Both run a named internal operating system — Danaher has the Danaher Business System, Thermo Fisher has what it calls the PPI Business System. Both have compounded revenue through relentless deal-making. But Danaher has repeatedly made itself smaller, spinning off Fortive in 2016 and Veralto in 2023, and now runs a more concentrated, higher-margin portfolio of life-science and diagnostics businesses.

Thermo Fisher has only ever broadened — and broadened into lower-margin services. The question this raises is uncomfortable and worth asking plainly: is Thermo Fisher a compounder, or a conglomerate that has not yet had its Veralto moment? A company whose highest-margin segment is 23% of revenue and whose lowest-margin segment is 56% is precisely the shape that eventually attracts someone arguing for a break-up. No activist has appeared. We note the setup without predicting it.

Thermo Fisher had an extraordinary pandemic. Reported COVID-19 testing revenue peaked at roughly $9.2 billion in 2021 — close to a quarter of everything the company sold that year — and then collapsed by about two-thirds the following year. Total revenue still rose 15% in 2022, because PPD and the core business absorbed a six-billion-dollar hole. That is a genuinely impressive piece of corporate absorption and it deserves saying.

Here is what happened next.

Read the third and fourth rows again.

Revenue in 2025 ($44.56bn) is still below revenue in 2022 ($44.91bn). Three years, no progress on the top line.

GAAP earnings per share in 2025 ($17.74) is still below 2021 ($19.46). Four years, and the accounting earnings of this company have gone backwards — despite roughly $30 billion of acquisitions and about $15 billion of share buybacks in the same window.

And free cash flow has not grown at all. It has sat between $6.3 billion and $7.3 billion every single year from 2020 to 2025. Six years, thirty billion dollars deployed, and the cash the business produces is no larger than when it started.

This is not a story about a de-rated compounder. It is a genuine earnings plateau. The multiple did not fall because the market lost its mind — it fell because there was nothing new to pay for.

Capital expenditure fell from $2.52 billion in 2021 to about $1.4–1.5 billion in 2024 and 2025 — a cut of roughly 40% — while revenue grew. That is either admirable discipline after a pandemic building spree, or it is under-investment. We cannot resolve it from outside. What we can say is that it flattered free cash flow in precisely the years the story most needed flattering, and that it is worth watching whether capex stays down now that the end markets have turned.

And the news is not all backward-looking. 2026 is an acceleration year. In the second quarter, reported 23 July, revenue rose 10% to $11.99 billion with organic growth of 5%, adjusted operating margin expanded 90 basis points to 22.8%, all four segments grew, and management raised full-year guidance to $47.4–48.1 billion of revenue and $24.93–25.33 of adjusted earnings per share. The end-market cycle has visibly turned. The bear case is now about the price and the balance sheet, not about the business breaking.

Everything above leads to one question. A company that grows by acquisition is only a compounder if it earns a good return on the money it spends. Revenue growth bought with capital proves nothing. So: what does Thermo Fisher earn on the capital it has deployed?

That is the finding, and it is worth stating without decoration. Thermo Fisher is not destroying value — earning your cost of capital is not failure. But a business reinvesting at roughly its hurdle rate is not compounding either. It is running to stand still on a larger and larger base.

Return on equity of 13.4% looks better, and it is the number most often quoted. It is higher only because the balance sheet carries $38.5 billion of net debt. Leverage flatters returns to equity; it does not improve the return on the enterprise.

Thermo Fisher reports two sets of earnings, and the gap between them is large.

Roughly 22% of the earnings presented to investors consists of costs the company asks you to look past — dominated by the amortisation of intangible assets created when it buys companies, with restructuring and acquisition expenses behind it. On about 380 million shares that is close to two billion dollars a year.

Now — what would Buffett actually say about that? This is where we have to be careful, because the lazy answer is wrong and we very nearly published it.

The lazy answer is that Buffett insists on the audited number and would reject the adjustment. He would not. In his 2015 letter to Berkshire shareholders he wrote, on precisely this point:

"For software, as a big example, amortization charges are very real expenses. Conversely, the concept of recording charges against other intangibles, such as customer relationships, arises from purchase-accounting rules and clearly does not reflect economic reality." — Warren Buffett, 2015 shareholder letter

He then did exactly what Thermo Fisher does: he presented Berkshire's own operating businesses on a basis that excluded most of that amortisation, telling shareholders that of Berkshire's own amortisation charges, only about a fifth were "real". On this narrow question, Thermo Fisher's management is standing where Buffett stands. We think the adjustment is defensible and we would not build a bear case on it.

But he drew a hard line in the same letter, and it is the line that matters here. On depreciation — the wearing-out of actual physical assets — he was unambiguous:

"When CEOs or investment bankers tout pre-depreciation figures such as EBITDA as a valuation guide, watch their noses lengthen while they speak." — Warren Buffett, 2015 shareholder letter

And, on the same subject: "I wish we could keep our businesses competitive while spending less than our depreciation charge, but in 51 years I've yet to figure out how to do so."

So the correct Buffett lens on Thermo Fisher is not "use the GAAP number". It is this: forgive the amortisation, because it is an accounting shadow — but never forget that the cash which created that shadow was entirely real. Thirty billion dollars left the building. The honest test is not which earnings figure you print. It is what return the company earns on the full amount it actually spent, goodwill and all.

That number is about 7.5%.

And it is worth noticing the smaller point too: this is a company whose capital expenditure has run at little more than half its depreciation and amortisation charge for three years running. Some of that gap is the intangible amortisation that Buffett would forgive. Some of it is not.

⚠️ One governance fact deserves stating, with its caveat attached. Casper's total compensation for 2025 was reported at approximately $80 million, including around $66 million of stock awards under a five-year cliff-vesting performance grant. We have not verified this against the proxy statement and it should be treated as a press figure rather than a confirmed one.

If it is right, the timing is worth a sentence. The board granted a very large equity award in the year the shares bottomed at $446 — and equity struck at a depressed price is enormously valuable if the stock merely returns to where it was, which it largely has. The fair counter-argument is that a five-year cliff-vesting performance award is genuinely long-term by American standards, and retaining the person who built the company has real value. Both things are true. What we could not establish, and what would settle the question, is which metrics the award actually pays on. If it rewards revenue and adjusted earnings growth, it directly incentivises the empire-building this report has spent two parts questioning. If it rewards return on capital, that is materially better governance. We flag it as unresolved rather than guess.

Our house rule is that any dividend-paying company gets its payout examined properly — coverage on free cash flow rather than earnings, the trend, and what would have to happen for a cut. Thermo Fisher pays a dividend, so here is that examination. It is short, because the answer is not close.

The dividend is safe to the point of triviality, and that is exactly the problem with treating it as a reason to own the shares. A 0.32% yield is a rounding error in the capital plan and a rounding error in your return. Even at 3.5 times net debt to EBITDA after Clario, the payout is not competing with anything.

For a reader who has followed our work on PepsiCo or Main Street Capital, the contrast is the useful part. There, the dividend was the investment case and the coverage analysis was the whole report. Here it is a signalling device and an index-eligibility box. If you buy Thermo Fisher, you are buying capital growth. Nothing else is on offer.

The premise we started with does not hold. A stock 16% below its high, up more than a fifth in a year, trading at its own ten-year average multiple, is not a quality company on sale. It is a quality company at its normal price. The discount existed in April 2025 and it has been arbitraged away.

And the balance sheet deserves its own line here, because it is the part the market pays least attention to. Goodwill and other intangibles together came to $73.4 billion in June 2026, against total assets of $113.2 billion and shareholders' equity of $52.8 billion. That is roughly 65% of the balance sheet, and it means tangible book value is about negative $20.6 billion: everything the company owns in physical reality — plants, inventory, receivables, cash — is financed by debt and other liabilities, and the equity account exists because of what it paid for businesses rather than what it owns.

That is neither fraud nor unusual for a serial acquirer, and we do not raise it as an accusation. We raise it because it means the cushion against a large impairment is thin, and because an impairment — though non-cash — would force a public reckoning with prices paid that the market has so far been content to leave alone. No impairment has been taken on PPD. That is management's assertion that the deal is fine. It is not independent evidence that it is.

A fair steelman of the bull case, because it exists. You are paying roughly 22 times forward adjusted earnings for the dominant supplier to an industry with genuine secular growth; about 42% of revenue is true razor-blade consumables; the crown-jewel segment earns 37% margins behind switching costs enforced by law; the end-market cycle has visibly turned with organic growth reaccelerating and guidance raised; and the litigation record is remarkably clean. That is a reasonable price for quality. It is simply not a bargain, and it offers no margin of safety. We score valuation a 4.