Meta’s $35 Billion Bet on AI Fuels the Tech Arms Race

AI and money

Meta announced a $35 billion investment in artificial intelligence (AI) for this year, signaling an aggressive push in the escalating tech arms race.

This colossal investment raises pivotal questions about the future of AI development and its financial viability. Industry experts are now debating the scope and impact of this funding, probing into when these investments will yield a return on investment (ROI) and how they could reshape Big Tech revenue models. With potential strategies ranging from subscriptions to advertising, Meta’s move could set new precedents for how technology giants capitalize on AI advancements.

“For now, there seems to be no end in sight to the arms race,” Muddu Sudhakar, CEO of generative AI company Aisera, told PYMNTS. “AI is clearly a top strategic focus. Think of it like the transition from on-prem to the cloud or from the desktop to mobile. These are massive secular trends that last many years. So, for a megatech company like Microsoft, Google, Meta, or Amazon, missing out on AI would be disastrous. This is why they are pouring billions of dollars into capex.”

Growing AI Investment

Meta’s recent earnings report revealed that instead of boosting its return on investment from AI, the company is increasing its spending by $5 billion to develop new AI products for consumers, developers, businesses and hardware manufacturers.

The company’s investment in AI and its metaverse development arm, Reality Labs, is projected to reach between $35 billion and $40 billion by year’s end.

CEO Mark Zuckerberg also discussed the launch of the latest version of Meta’s AI assistant, Meta AI, last week, which is enhanced by the newest updates to its large language model, Meta Llama 3.

Despite the massive investment in AI for Meta and other companies, observers say the returns on investment might be far off.

“We are in the investment stage of the AI cycle,” Sudhakar said. “So it’s unrealistic to expect wide-spread ROI. Another historical example is the early days of the Internet. There had to be experimentation with use cases, education, adoption, major investments in infrastructure, and so on. The same thing is happening with AI. Although, there is one area where there is clear ROI: the infrastructure providers, especially Nvidia. But over time, the ROI will broaden out, such as to applications.”

Costs May Hold Back Profits

While the excitement and investment in artificial intelligence is building, the technology — especially generative AI — is very costly, Sudhakar noted.

“There are the costs for GPUs, data science talent, data center build-outs, energy usage, data management,” he said. “But the good news is that the costs will start to decline, which will help with monetization. For example, this week, Snowflake announced its own LLM. It cost only $2 million to train and required 1,000 GPUs.”

The value of AI comes not just from products and services sold to end users but also from internal company efficiencies and savings, Lars Nyman,  chief marketing officer at CUDO Compute, told PYMNTS.

“Granted, some applications, like fraud detection, chatbots, or better ad serving algorithms deliver clear, direct financial returns fast,” he said. “But for fundamental research projects, the ROI might be fuzzier, measured in long-term innovation advantage — measured over years, not quarters.”

According to Sudhakar, Big Tech companies are likely to adopt a blended business model to profit from their AI models. Initially, those with substantial cloud infrastructures can capitalize immediately by monetizing their investments through a usage-based model for cloud services. This often involves accessing large language models (LLMs) and smaller language models (SLMs) via an API. Additionally, there are opportunities to cross-sell development tools and various other services within this ecosystem.

Furthermore, as Sudhakar noted, AI applications can be marketed as subscriptions. For instance, OpenAI has seen considerable success with its ChatGPT system, while Microsoft has achieved similar success with GitHub Copilot and is exploring possibilities with its Office 365 Copilot offering. Besides subscriptions, advertising presents a viable revenue stream, particularly for companies like Google and Meta, which can leverage their extensive advertising infrastructure and AI enhancements to optimize ad revenues.

“However, Google is in a tough spot,” he said. Its core search business relies heavily on getting paid for links. But if a generative AI app can provide much of the information in one response, there will be pressure on this revenue stream.”

Yigit Ihlamur, general partner at Vela Partners, told PYMNTS that once advertising is introduced to AI, revenues will skyrocket.

“For example, if you use ChatGPT to book a flight, organic and sponsored results would show up,” he said. “I imagine tech companies would further monetize their models with paid subscriptions, developer-focused paid APIs, and marketplace transactions.”

 For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.