By Clare Pleydell-Bouverie and Storm Uru, fund managers on the Liontrust Global Innovation team
If there has been one burning question on investors’ lips this earnings season, it has been when will AI start to generate meaningful revenues? The $210bn in capex slated for 2024 from just four of AI infrastructure’s biggest spenders (Alphabet, Meta, Amazon and Microsoft) justifies the question.
These companies and many more are investing heavily in AI and plan to increase investment further still in 2025. Why? Sundar Pichai, CEO of Alphabet, summed it up perfectly on the company’s second quarter earnings call: “The risk of under-investing is dramatically greater than the risk of over-investing for us here.”
Two key points are packed into this sentence, the sentiment of which was echoed across Silicon Valley this earnings season. The first is that the expected gains from AI are worth pursuing, and this is based on evidence from early AI use cases already driving compelling returns on investment.
See also: Nvidia share price drops in after-market trading despite outperforming forecasts
The second is that this investment risk can be managed through sufficient cashflow generation. Microsoft, for example, has spent a whopping $44.47bn on capex over the past 12 months (up 60% year-on-year), yet this has left 63% of operating cashflow untouched.
So where is this elusive return on investment (ROI)? You will likely not find it in the financial statements of the majority of companies touting their AI credentials on earnings calls.
Artificial intelligence has certainly become a buzzword, but that does not mean that select pioneers in AI deployment are not witnessing outsized gains.
The AI pioneers
Meta stands out among the tech giants, with the company’s AI investment in content recommendation and AI ad-tools already driving tangible revenues. Over 50% of the content we see on Instagram is now recommended by AI, while advertisers using Meta’s AI advantage+ tools are seeing a 22% higher return on ad spend.
These factors helped power advertising revenues 23% higher year-on-year in the second quarter. This is the ROI that everyone is looking for, there are just very few companies today capable of deploying AI at scale and generating meaningful revenues.
See also: Choosing ‘picks and shovels’: Managers look beyond Nvidia for AI plays
ServiceNow, the leading enterprise software company for workflow automation, is also in this select minority of companies deploying AI at scale today. After using Now Assist, the company’s Generative AI assistant, internally for 90 days, the company realised over $10m annualised enterprise savings.
Customers using the Gen-AI tool are seeing similarly dramatic productivity gains – BT Group has cut the time it takes for its agents to write and review cases by 55%. Now Assist has become the fastest growing product in the company’s history and customers are paying up.
Beyond big tech
When it comes to smaller companies, meaningful AI-related revenues are more elusive but can be found. These include Palantir, whose AI-powered platform, AIP, enhances data analysis and operational efficiency across corporations and defence organisations. Using AIP, General Mills have driven $14m annualised cost savings. Demand for Palantir’s AI solutions is not only propelling its topline growth (up 30% year-on-year in the second quarter) but has turned the business profitable over the past seven quarters.
One thing that all these companies have in common is a world class compute infrastructure. Meta, for instance, trained its most recent large-language-model, Llama 3, on two GPU clusters each comprising 24,000 Nvidia H100s. Llama 4 will be trained on 10x this compute. This is the infrastructure that is necessary to deliver the next step up in model performance, which will enable further revenue opportunities.
Not only do we believe this investment is justified, we believe it is rational capital deployment. For every $1 invested by cloud service providers in Nvidia-accelerated computing infrastructure (the GPUs and related kit required to build and train AI), this will translate to an estimated $5 over four-years.
See also: The tech giants are facing an ‘existential event’
This is because AI solutions are bringing customers to the cloud, driving revenue growth for providers, which will translate to returns on their infrastructure investments. For companies serving AI models (essentially providing the kit for AI inference), the ROI is even more significant – $1 invested today has the potential to generate $7 in revenue over a four-year time horizon. It is this compelling ROI on AI infrastructure that is leading companies to accelerate their investments in AI infrastructure, despite recent investor consternation.
AI-generated revenues will not be enjoyed by all. AI is a platform technology shift that, like all superseding platform transitions, will entail disruption as well as opportunity. Furthermore, companies have to build this infrastructure before they can monetise it.
As such, we believe that the value for investors today lies predominantly in the AI infrastructure layer of the new technology stack. Over time the baton will get passed to the AI application layer.
We expect this baton to be passed on more quickly than in previous technology shifts, because the gains have potential to be much larger. With the cloud transition it took Microsoft seven years to reach $3.5bn annualised revenues – with generative AI the company achieved this milestone in just 18 months.
