Broadcom Doesn't Innovate Into New Markets. It Buys Them, Then Stops Spending.

In 2015, a chipmaker named Avago agreed to buy a chipmaker named Broadcom for $37 billion — and then did the one thing nobody expected. It kept the other company's name and threw away its own.¹ The combined business, worth roughly $77 billion in enterprise value, began trading as Broadcom in February 2016.² Most people remember this backwards: they think Broadcom bought Avago. It was the reverse. And that small, deliberate inversion — the acquirer wearing the acquired company's face — tells you everything about how this company grows.

The official story is that Broadcom is a semiconductor innovator that branched into software. The truer story is that Broadcom is a financial machine wearing a technology company's clothes. It does not branch. It buys. And then it does something the engineers who built those products never imagined: it stops spending on them.

The playbook is the product

Here is the thesis a smart friend could repeat at dinner: Broadcom does not expand into adjacent markets by inventing things for them. It acquires mature, cash-generative businesses with customers who can't easily leave, cuts the research and development that those customers were never paying for anyway, and re-prices the relationships that remain. The genius isn't a product. It's an arbitrage — between what a sleepy enterprise asset earns under patient ownership and what it earns under owners who treat R&D as a cost to be minimized rather than a future to be funded.

Watch the moves in sequence and the pattern stops looking like a strategy and starts looking like a template. In July 2018, Broadcom agreed to buy CA Technologies — the old Computer Associates, a mainframe-and-enterprise-software relic with deeply embedded customers — for $18.9 billion, closing that November. It was the first real pivot out of chips and into enterprise infrastructure software.⁶ CA made nothing exciting. That was the point. It made recurring, predictable, sticky revenue from clients who had no realistic way to rip it out. Then came the big one: VMware.

Why cutting R&D is the feature, not the bug

Most acquirers buy a company for its future — its roadmap, its next product, its growth. Broadcom buys a company for its present, and specifically for the part of the present that the prior owners were quietly subsidizing. A mature enterprise software business spends heavily on research because that's what its culture says a technology company does. But its customers — banks running CA on mainframes, enterprises that built their entire data centers on VMware — are not buying the roadmap. They're paying to keep the lights on. They are, in the most literal sense, locked in: the switching cost of migrating off the platform dwarfs almost any price increase you could put on it. So the R&D line is, from a cold financial seat, money spent winning customers who were never going to leave.

Strip that spending out and two things happen at once. Margins jump, because a large cost disappears. And the revenue barely moves, because the customers can't go anywhere fast. Then re-price the contracts upward, lean on the lock-in, and you have converted a respectable software business into a cash geyser. The asset didn't get better. The owner just stopped paying for a future the customers weren't buying — and started charging more for the present they couldn't escape.

The deal the government read out loud

The cleanest proof that this is the actual playbook didn't come from a critic or a short-seller. It came from the United States government, which described the strategy in writing and then used it as the reason to kill the largest deal Broadcom ever attempted. In November 2017, Broadcom made an unsolicited bid for Qualcomm — the crown jewel of American wireless — at $103 billion, later raised to $117 billion.⁵ It would have been the biggest tech takeover in history. It never happened.

On March 5, 2018, the Committee on Foreign Investment in the United States delivered a national-security impact assessment. Its logic is worth reading slowly, because it is the Broadcom thesis stated by an adversary: CFIUS cited Broadcom's own prior statements that it would reduce Qualcomm's long-term R&D investment to boost short-term profitability — and warned that doing so to a company at the center of 5G would leave an opening for China to expand its influence over 5G standard-setting.⁴ A week later, on March 12, 2018, a presidential executive order permanently blocked the deal on the recommendation of CFIUS.³

Read that again. The government did not block Broadcom because the playbook was a fantasy. It blocked Broadcom because the playbook works — works so reliably that regulators could predict, in advance, exactly what would happen to Qualcomm's research budget the morning after the deal closed. The strategy that minted enormous value on CA and would later mint it on VMware ran into the one asset where cutting R&D wasn't merely the buyer's prerogative but a matter of national strategy. Same machine. The wall it hit was the point.

Isn't this just good capital discipline?

The fair objection is that none of this is cynical — it's simply rigorous capital allocation, and the world has too little of it. Plenty of enterprise software companies do spend on R&D that goes nowhere, defending roadmaps no customer asked for. If Broadcom is willing to make the unsentimental call and redirect that capital to shareholders, isn't that exactly what a public company is supposed to do? There's real force here. Bloated incumbents that confuse activity with progress deserve an owner who counts. And Broadcom did clear regulators in more than a dozen countries to close VMware — including the EU, the UK, and China — which is not the record of a pure asset-stripper.⁷

But the honest answer is that discipline and extraction look identical in the quarter and diverge over the decade. Cutting R&D that funds nothing is discipline. Cutting R&D that funds the future your locked-in customers will eventually need — and trusting that they can't leave fast enough to punish you — is a tax on captivity. CFIUS named the moment the two stop being the same thing: when the asset's research isn't slack to be trimmed but infrastructure a country is standing on. The playbook is brilliant precisely where the customers can't leave. Its ceiling is the place where someone with more power than the customers — a government, a standards body, a rival platform — decides they shouldn't have to.

Broadcom expanded beyond its core the way a holding company expands — by purchase, not by invention, and by the patient conviction that a great enterprise asset is worth more to an owner who stops dreaming about it. Avago bought a name in 2015 and wore it; it bought CA and VMware and ran the same machine on each. The strategy isn't a secret. The U.S. government wrote it down. And in writing it down to stop the Qualcomm deal, it accidentally certified the whole thesis: the playbook is real, it is powerful, and it works right up until it meets something its customers can't be made to swallow. The genius was never a product. It was finding the rooms nobody could walk out of — and the only question that has ever mattered for Broadcom is whether anyone with more power than the people inside is watching the door.