As the AI data centre buildout gathers speed and increases spend, the hyper-scaler companies are changing in important ways that investors need to note.
The hundreds of billions being spent is collapsing cash flow and changing their balance sheets from asset light to early days of asset heavy.
Selling an advert against a search query or Facebook rant required very little actual computer power sitting in a data centre and, as such, margins and free cash flow were huge. This was equally true for Microsoft selling software that has little ongoing costs after the initial production cost.
Recent results show that not only is free cash flow under pressure, in some cases it is gone. Alphabet reported negative free cash flow per share for the first time since it listed in August 2004.
On the debt side Meta’s long-term debt has surged to $83.7bn against $90.3bn in cash. Its results also showed operating margin collapsed by almost a third — though that operating margin is still 31%, a strong number. But operating profits fell even as revenue was up 28%.
The move to debt is a sudden issue of bonds and, in the case of Alphabet, even selling some stock to raise capital. This hurts longer-term profits as it needs to pay the interest on the bonds, but the numbers here are not yet big enough to move the needle. Not when compared with the hundreds of billions being spent on capital expenditure over this and the next few years.
There are also other implications. Share buybacks are always funded from free cash flow and both Meta and Alphabet did zero buybacks in the latest quarter. Again, not immediately material. But it does remove a buyer from the market and often these bought-back shares would give the companies space to issue shares to staff instead of compensation. They can still do that, but now it will push the share count higher.
There is also the depreciation story. Capex is not an expense when you spend it, but it will come out through the P&L (profit and loss account) in the years ahead and there is a wall of capex spending that will have to exit. It will be noncash when it gets charged, but there is plenty opportunity to spook the market.
Ultimately, if the income arrives as expected (and expectations are huge) then the spending will have been worth it.
It is still early days and confidence remains high about the potential. But as investors we have to get used to viewing these companies differently as they evolve into AI giants with balance sheets with both debt and lots of assets. They will also have to act differently. Alphabet, for example, has less space for its side projects and Meta’s Reality Labs will have harder questions being asked as revenue remains a missing link.
Ultimately, there are a lot of unknowns here, but at the heart the one known is that these tech companies are changing and so is the investment story.