What's ahead for big tech?

Tech, power, and some other stocks were jolted on January 27 when a Chinese generative artificial intelligence (AI) company released the latest version of a free, large-language model (LLM). Many mega-cap tech stocks—some of the Magnificent 7 in particular—had their worst day in years. Since then, most of the hardest hit stocks recovered some of their losses.

Here’s what investors need to know about the latest AI news and how it might impact them.

Artificial intelligence competitor emerges

A foreign startup launched its V3 LLM back in December. On January 27, the same company released its new R1 model (which was built upon the V3 model) and quickly became a top downloaded app.

The company reported that the open-source LLM reportedly cost just $5.6 million and it appears to be functionally competitive to other AI tools.¹ That being said, that figure is related to the last training run of the model, and we don’t have a full accounting of the entire spend picture. In addition, the cost doesn’t incorporate the costs associated with the open-source LLM models that were used to train R1.

Speed to market was another factor that took some investors off guard. Instead of taking months or years to pretrain the LLM, as has been the case with other models, R1 reportedly took just 2 months. Again, it is worth noting that the model was built using more reinforcement learning—building on top of existing models—making it faster to train and more efficient than one with no previous versions.

The news reinvigorated conversations about AI spend and within the AI community about whether algorithmic efficiency, represented in this case by the R1 LLM, may be a more efficient approach than brute-force computational power. The latter has been a driving factor behind massive spending by hyperscalers on big-data centers, chips, and power generation.

Investing implications

The R1 model raised questions about the estimated cost of building out AI going forward. Due to the compute costs for the R1 LLM appearing to be a fraction of that of other large foundational model builders, semiconductor and hardware stocks (which have far outperformed the broad market over the past several years) were rattled upon the news. Some power stocks, which have benefited from computational energy needs, were also impacted.

While it’s still early, there was some rotation into software stocks from semiconductors, given the potential implication that new models could potentially be built on the backs of existing ones using more software and less hardware-intensive processes.

So, what does this mean for investors? Pri Bakshi, a portfolio manager at Fidelity, thinks this could have positive implications on AI adoption going forward, as investors may see more take-up in AI if it’s less expensive. However, he notes that we’re still very early in our understandings of AI, and we can expect the landscape for AI investing to continue to evolve over time.

Adam Benjamin, another portfolio manager at Fidelity, notes companies finding more efficient ways to train LLMs isn't a surprise, and investors should continue to expect companies to search for efficiencies. While some stocks sold off on concerns sold off on concerns that the models would require less compute, Benjamin thinks those companies can still offer value. R1 could accelerate the move from training to inference (i.e., using trained models to analyze new data and make predictions), which could drive a substantial increase in compute requirements as we are still in the early phases of the rollout of AI and its use cases.