Broadcom Doesn't Buy Software Companies. It Buys Captive Customers.

Less than a month after Broadcom finished buying VMware, the bill arrived. The perpetual licenses that companies had paid for once and owned forever simply stopped existing; everything moved to subscriptions, the product menu was gutted and reshuffled, and the renewal quotes that landed in IT inboxes were not a little higher - they were 2x to 12x higher, depending on what you ran.⁷ These were not fringe customers. Roughly 80% of organizations run VMware infrastructure, some 375,000 of them worldwide.⁷ Broadcom had not bought a software company. It had bought a population that could not leave - and then it raised the rent.

The official story is that Broadcom is a chipmaker diversifying into software. That framing misses the entire point. Broadcom is not diversifying. It runs one move - the same move - over and over, on whatever it can get its hands on. The chips are real, but the software is where the discipline shows.

The thesis: this is a toll road, not a portfolio

Here is the whole strategy in one sentence. Broadcom acquires enterprise software that holds a near-monopoly position over customers who are locked in by switching costs, strips the cost of running it down to the bone, and then raises the price on people who have nowhere else to go. It is not a bet on growth. It is the acquisition of a captive audience, repeated. The VMware deal closed in November 2023 at an announced equity value of about $61 billion in cash and stock - roughly $69 billion once you add the $8 billion of net debt Broadcom assumed.¹ Before it, there was Symantec's enterprise security business for $10.7 billion in 2019, which Broadcom itself called 'the next logical step in our strategy following our acquisitions of Brocade and CA Technologies.'² Notice the word: strategy. Not opportunism. A repeatable mechanism.

The part where the cost gets cut

The cutting is not a regrettable side effect. It is the engine. When Broadcom closed its $18.9 billion purchase of CA Technologies in 2018, it laid off 300 inherited employees almost immediately and announced it would cut about 2,000 more - out of a workforce of roughly 4,800.⁵ That is not pruning. That is removing the better part of the company you just paid for, because you did not buy it for the company. You bought it for the recurring revenue stream the company happened to be sitting on. The same pattern ran through VMware. WARN notices filed for the Palo Alto campus alone covered 1,267 employees, with hundreds more in Georgia and Colorado, against a VMware that had more than 38,000 people the prior February.⁶ The math is brutal and deliberate: the revenue stays, the people don't, and the difference falls to profit.

The government wrote the playbook down before Broadcom ran it

The most damning evidence for what Broadcom actually does is that a U.S. national-security panel predicted it in writing - and then Broadcom went and did it to three other companies. In March 2018, President Trump blocked Broadcom's hostile bid for Qualcomm on the recommendation of CFIUS, which found that Broadcom's stated intent to slash Qualcomm's long-term R&D would leave an opening for China to dominate 5G standard-setting.⁴ Strip away the geopolitics and you are left with a regulator's clean diagnosis of the model: this buyer cuts the future to harvest the present. That is precisely what followed at CA, at Symantec, and at VMware. CFIUS wasn't wrong about Broadcom's intentions. It was just describing them a few years early.

The numbers leave little doubt that the machine works. In fiscal 2024, infrastructure software revenue rocketed 181% to $21.5 billion - almost entirely VMware - dragging total revenue up 44% to $51.6 billion.³ A year later, adjusted EBITDA hit a record $43.0 billion, free cash flow reached $26.9 billion, and the CFO guided to a 67% adjusted EBITDA margin.⁸ A two-thirds EBITDA margin is not what a software business earns by delighting customers. It is what a toll road earns from drivers who have no other route.

Isn't this just disciplined capital allocation?

The honest objection is that this looks less like predation and more like grown-up management. Plenty of acquired companies were bloated; plenty of perpetual-license software was underpriced relative to the value it delivered. If customers genuinely couldn't function without VMware, the argument goes, then VMware was simply charging too little, and Broadcom is correcting a market failure while funding the AI-chip business that grew 74% year-over-year.⁸ There is real force in that. Capital does flow to where it compounds, and a 67% margin is the market voting that the assets are worth more in Broadcom's hands than where they were. But the steelman quietly concedes the thesis. The reason Broadcom can reprice at 12x and still keep the revenue is not that the product got 12x better - the product menu got narrower. It is that the customer is trapped. A strategy that depends on the buyer having no alternative is not a story about creating value. It is a story about capturing it from people who can't say no. The two can wear the same financial statement, but they are not the same thing - and the difference is exactly what CFIUS flagged.

Broadcom makes its money the way a tollbooth does - indifferent to where you're going, certain only that you have to pass through. Buy the road every enterprise already drives on, fire the people who built it, and charge each driver what it would cost them to build their own. CA proved the model, Symantec scaled it, VMware made it enormous, and Qualcomm's regulators described it before any of it happened. The genius was never the engineering. It was recognizing that the most valuable thing a software company owns isn't its code - it's the customers who can't afford to leave it. And every time one of them grits their teeth and signs the renewal, a little more of that value moves to Broadcom's bottom line.