Tech stocks ask for patience on AI revenues

So far, the artificial intelligence (AI) stock market rally has been built on faith and fear. Faith that generative AI will be productivity-enhancing, widely adopted and valuable to end users, and the fear of being left behind, by both competitors and those investors reaping the early gains. The immediate consequence has been a huge ramp-up of capital spending on AI servers. What we are yet to see is a software product that generates significant revenue and returns on all this investment.

Cloud computing companies have been racing to buy graphics processing units (GPUs), the hardware needed to train AI models. This is why Nvidia (US:NVDA) saw its revenue double year on year in its most recent quarter. So far, however, that is the only place in the AI supply chain where there has been meaningful revenue uplift.

This year, the cloud computing companies have all cut operational costs to boost their capital expenditure. They are yet to see this investment pay off. For now, it is just an additional drag on their return on equity (ROE). As the graph shows, between December 2021 and June 2023, Alphabet’s (US:GOOGL) ROE has fallen 8.7 percentage points to 23.3 per cent, according to FactSet. Similarly, Meta’s (US:META) is down 13.7 percentage points to 17.3 per cent, while Microsoft’s is down 10.2 percentage points to 38.8 per cent.

A declining return on equity is expected when companies invest so much in such a short space of time. Eventually, you would expect ROE to start improving, assuming customers eventually start paying up for the new improved products. However, public markets are historically impatient during these transitions.

 

Painful transitions 

We saw this when software companies first started their move to the cloud. That required a lot of upfront costs to redesign products, and then time to convince customers to switch to their software-as-a-service product. It was a lengthy process, which the UK market in particular became frustrated with.

Accounting software company Sage (SGE) saw its operating profit slip from £411mn in 2020 to £358mn the following year as it ramped up its research and development spending. The share price lagged as investors questioned whether this would pay off. However, this year those who were patient have been rewarded. In the six months to March, Sage published double-digit annual recurring revenue growth and an expanding operating margin. This year, its share price is up 37 per cent.

It was a similar story at UK-listed industrial software company Aveva, albeit with a different ending. In 2022, Aveva’s profit margin fell because it decided to bring forward £20mn of R&D spend to speed up the transition to the cloud. It told the market the plan. Broker Numis said it was modelling an earnings “J-curve”, in other words, a dip before a big rise. But the share price still fell. Eventually, Aveva lost patience with its valuation and sold the rest of the business to its main shareholder, Schneider Electric.

The issue for software companies is their end users are resistant to change. People get used to a certain way of working and it can be painful to adjust. This dynamic is still playing out in the cloud transition, which is far from complete; according to Ofcom, less than 15 per cent of IT spending is on the cloud.

 

AI revenue barely in sight 

Microsoft (US:MSFT) is one of the software companies closest to receiving a meaningful revenue uplift from AI. It is charging $30 (£24.50) per user per month to use its AI-enabled Office 365 ‘Copilot’. This is additional to the $36 per month customers currently pay for the regular Office 365, equivalent to an 83 per cent price increase. It’s expensive but Microsoft sounds confident this price is reasonable. “Feedback from organizations like Emirates NBD, General Motors, Goodyear and Lumen is that it’s a game changer,” chief executive Satya Nadella asserted on a recent earnings call.

However, despite this insistence, Microsoft is being careful when it comes to the hard numbers. In its July earnings call, the company forecast that meaningful revenue growth from AI will only start appearing in the second half of 2024. “Even with strong demand and a leadership position, growth from our AI services will be gradual,” said chief financial officer Amy Hood.  

The market paused to see if the AI story into which everyone has bought will actually start producing real cash flows. After a big run-up at the start of the year, Microsoft’s share price has been flat since July. As Jefferies put it: “Investors have more realistic expectations for FY2024 revenue contribution related to AI.”

At the other large software providers, AI revenue is even further away. Adobe (US:ADBE) is using its AI product as more of a marketing ploy. In September, it launched its text-to-image generative AI product Firefly, but has made it free to use via a web app. Adobe referred to this strategy as “building the top of the funnel”. The hope is if lots of people start using the free product, some of them will eventually upgrade to a paid-for subscription.

The concern for investors is that these AI models are expensive to build and expensive to run. Every time someone uses Adobe Firefly to generate an image there are computing costs and so far Firefly has created more than 2bn images.

To maintain its margin, Adobe has had to boost profitability elsewhere in the business. When asked by Morgan Stanley analyst Keith Weiss how it was possible that Adobe could be offering Firefly for free but maintaining its margin, chief financial officer Daniel Durn kept it vague by just praising “operating philosophies”.

Weiss subsequently predicted there “will likely be a gross margin impact on a go forward basis given the costs to generate every Firefly image”. However, he gave Adobe the benefit of the doubt, saying it “sounded confident in its ability to continue to find operational efficiencies”. The thing about operating efficiencies is they come from increased scale. Firefly will need to keep bringing customers through the funnel if it is to hit this target.

Salesforce (US:CRM) is similarly far away from generating any revenue from AI; it won’t do so for at least a year. The company has just released an AI product called Einstein Copilot. However, analysts don’t forecast any revenue uplift until 2024 or 2025. “[Salesforce] partners noted they have received inbound interest from customers for educational meetings around AI, but that has not translated to orders,” Jefferies’ tech analyst Brent Thill said.

Unlike Adobe, it is obvious how Salesforce has managed to maintain its margins. At the start of the year, it laid off 8,000 employees, equivalent to 10 per cent of its workforce. To keep investors on side until these AI revenues appear, it needs to maintain its profitability. The longer it takes for AI to start producing revenue uplift, the more jobs may come under threat.

Jefferies analyst Mark Lipacis recently considered the question of whether Nvidia might suffer a slowdown in demand if an obvious use case for AI doesn’t emerge quickly. “There is a risk,” he said. “Competition is driving capex and there could be an overspend cycle right now. But there are already legitimate business models”.

The specific business model Lipacis had in mind is Meta. The difference between Meta and the enterprise software businesses is the former doesn’t need to convince customers to buy its product. Meta doesn’t have to go to market with its AI; instead, it uses it internally to boost advertising returns for its customers. So far, it’s been effective: last quarter Meta’s operating revenue was up 12 per cent year on year to $9.4bn. This is even more impressive given the advertising market is in a downturn, which has hurt the big traditional advertisers and crippled profits at smaller UK agencies such as S4 Capital (SFOR) and M&C Saatchi (SAA).

Given the amount of money cloud computing companies have spent on AI servers, they are clearly confident it is an essential part of our future. But these technology transitions are always messy. Even though companies have warned investors to be patient, there will be doubts along the way. Now the initial euphoria is past, investors will only truly believe when the revenue is there for all to see.