The subsequent stage of the AI boom

According to Terry Tanaka, the technology will soon be much more widely used

This is how to make money.

For a short while in January, it appeared that the US's technological hegemony was coming to an end. An AI start-up from China called DeepSeek quickly became the most downloaded app in the US app store. Not only was DeepSeek well-liked, but it was also capable of matching OpenAI and Google's rivals. According to its creators, it could accomplish this for only £6 million in training expenses and without depending on the state-of-the-art Nvidia chips that have prompted hundreds of billions of dollars in capital expenditures (capex) in the West.

That money appeared to have been spent very poorly, and Nvidia, a market darling that is priced for perfection, appeared to have lost the rationale behind its valuation of more than £3 trillion. The largest valuation decline in stock market history occurred when Nvidia's market capitalization dropped £589 billion in a single day. The majority of these losses were recouped in a matter of weeks, though Nvidia's stock has since dropped even more, which is largely due to the White House's erratic trade policies rather than concerns about AI's future.

The AI boom is still going strong. The light is going down, but Nvidia's day is far from over. "You have these secular themes that emerge every so often," acknowledges Michael Seidenberg, Allianz Technology Trust's lead portfolio manager. Secular in this sense denotes a long-lasting trend that involves excessive spending over many years. He asserts, "I don't doubt that AI is a secular theme." While DeepSeek hasn't put a stop to this, it has sparked a change in the AI cycle. Given this, it may be time to adopt a slightly different perspective on investing in AI.

Is Nvidia the top chip manufacturer still?

In terms of increasing the effectiveness of inferencingthe process by which a machine uses what it has learned to analyze new dataDeepSeek was unquestionably very inventive. However, there is some question if its grandiose claims are supported. Although it maintains that it did not use the most cutting-edge Nvidia GPUs (electronic circuits) available, it must acknowledge that this is the case because Chinese companies are officially prohibited from obtaining these chips. According to reports, the US Department of Commerce is looking into whether or not the model was created using restricted chips.

Asking which chips DeepSeek used or didn't use omits the point from an investment perspective: AI companies, including chipmakers like Nvidia, benefit from declining production costs. Nick Dumas-Williams, a technology investment analyst at Polar Capital, asserts that there are virtually no disadvantages to AI's decreased cost. Technology is naturally deflationary, and its uptake tends to increase as production costs decrease.

Increased volume should more than offset any direct pricing power that semiconductor companies like Nvidia or TSMC may lose. There is no reason to believe AI will take a different route given the skyrocketing revenue from cloud services and the declining cost of service provision. Furthermore, Daniel Ives, an analyst at Wedbush Securities, stated that Nvidia is still "leading the AI industry" and that there is a greater demand for its Blackwell chips than its supply. This year, Wedbush predicted that the Magnificent Sevens capital expenditures would increase from £100 billion to £325 billion. Despite the fact that any predictions made before the tariff disruption are now pointless, AI is still "the highest return-on-investment project that they can spend on," according to Dumas-Williams.

Even better performance is promised by the upcoming Rubin and Feynman generations of Nvidia chips. AI math continues to favor Nvidia as well. According to Dumas-Williams, "the AI scaling laws remain intact." "AI performance can be enhanced by increasing its processing power.

Even though that relationship is valid, AI developers are always motivated to increase their GPU spending. Nvidia, meanwhile, is examining the upcoming democratization wave. It's no accident that Nvidia's CEO, Jensen Huang, has begun to promote the company's PC products with the same fervor as its data center hardware. Once the economics reach the proper tipping point, a significant redesign of PCs is probably in the works. More markets mean more demand and expansion.

Is it still wise to invest in chip stocks?

On the flip side of the ledger, though, is the possibility that Nvidia's years of disproportionate share-price returns may be coming to an end due to the company's rapid expansion over the previous two years. Its yearly growth is decreasing and will eventually reach a plateau. That does not imply that the stock is bad or unworthy of being owned. However, it is unlikely to generate the same level of growth in the upcoming two years as it has in the previous two. New corporate applications of AI will be the focus of the next cycle's explosive spending growth.

Following DeepSeek, AI is becoming more affordable and transitioning from what Seidenberg refers to as the "test-and-develop" stage to "production." AI adoption by businesses is quickening, and they are beginning to notice improvements in their output and financial results. Consider Klarna. In September, it was able to terminate its Salesforce subscription after developing internal AI-replicating systems. After accounting for the 700 full-time contractors it also succeeded in replacing, Klarna saved £40 million annually.

Then there is Shopify, which instructed managers to demonstrate that AI is incapable of performing the job before they can hire a new employee. These kinds of actions will have a significant impact on businesses' bottom lines, particularly as agentic AI modelswhich make decisions and carry out more complicated tasks than merely visualizing or writingbecome more advanced.

In this regard, DeepSeek was a major driving force. It essentially invented the "distillation" technique, which uses a large, potent AI modelin this case, Metas Llamato swiftly and affordably train a smaller, more specialized model. Most importantly, for it to function, the "teacher" is necessary. For the large chipmakers, that alone ensures the demand for computers in the future.

We are about to witness a new generation of AI boom winners, but there is no reason to give up on investing in Nvidia just yet. The last phase saw the emergence of chipmakers, AIs, picks, and shovels. This is frequently described as a transition from one phase to the next. The software companies that create the tools that businesses use to enhance their performancewhether through cost reduction, productivity gains, or bothwill receive more attention in phase two.

This is comparable to the emergence of cloud computing, according to Anthony Ginsberg, CEO of GinsGlobal Index Funds. Alphabet, Microsoft, and, most importantly, Amazon were the first companies to win for cloud computing. "But the software-as-a-service (SaaS) guys are the bigger players once the initial infrastructural plumbing is in place," he claims. Here's a warning, though: finding winners is more complicated than simply searching for SaaS companies with AI capabilities. The AI boom might have some side effects, like any other revolution.

Over a decade has passed since Salesforce, arguably the quintessential software as a service (SaaS) company, went "AI-first" at the urging of its CEO, Marc Benioff. It now faces the possibility of being exposed. Therefore, what constitutes a quality second-wave AI play?

Next big winners of the boom: AI as a service.

AI is fundamentally a numbers game. Improved AI is a result of better training data. Therefore, through prudent mergers and acquisitions, Seidenberg recommends seeking out companies that have access to high-quality data sets." One that catches his eye is HubSpot, a company that creates software for sales, marketing, and customer support. A conversation intelligence platform called Frame AI and a corporate data platform called Clearbit are examples of "technology tuck-ins" that make up the majority of its 14 acquisitions. Specifically, Clearbit provides a wealth of data for a customer relationship management platform like HubSpot to mine, learn, and utilize. In contrast to some other businesses "that bought ten or twenty different companies and never unified the data model," it might give Hubspot "a homogeneous data set that really speaks to the use of AI," as Seidenberg puts it.

Another strategy is to find businesses that can incorporate AI into an existing product in a simple and efficient manner. As an example, Ginsberg cites EssilorLuxottica's Ray-Ban and its collaboration with Meta: "if you've got a platform or an existing brand that you can leverage with AI that, for me, is the software-as-a-service of AI," he says.

More generally, AI may have the greatest impact on industries with the most data and the most difficulties in converting it into a format that can be used. One such is medicine. Specifically, drug discovery is "the great white hope" of AI, according to Dumas-Williams. AI is already having a significant impact in a few specific instances.

Using DeepMinds AlphaFold, millions of protein structures are being mapped and their potential interactions are being predicted. Researchers are using it all over the world to study ways to treat osteoporosis, Parkinson's disease, and malaria. However, scale is the problem. Since a lot of research is conducted at the individual level, it is important to scale that up and integrate it into bio-research workflows. and takes so long to complete that I don't think we'll begin to realize this until the next few years," Dumas-Williams says.

How to make investments in the AI boom's next stage.

During that time, AI is probably going to help other data-intensive industries advance significantly. Among these, roboticsincluding driverless carsand the financial industry, with its robo-advisers, are at the top. It is impossible to predict which businesses will benefit from this, so diversify your investments in AI. AI ought to be widely adopted by most businesses. It's less clear that Nvidia will provide every chip needed to support that adoption or that any particular developer will produce the most revolutionary model.

This year has also seen two notable tech sell-offs, underscoring the risks associated with over-concentration. "The Seven Greatest. have consumed all of the oxygen present in the space," Ginsberg claims. The fact that many active strategies and market cap-weighted strategies focus on them "gives tech investing a very narrow view."

(LSE: ITEK) Ginsberg is the manager of the HAN-GINS Tech Megatrend Equal Weight UCITS ETF. The appeal of this exchange-traded fund is largely due to its equal weighting; assets are distributed equally among its 120 holdings, which reflect a fairly even spread across ten industries, including social media, cloud computing, robotics and automation, and future automobiles, all of which are primarily related to the AI theme. Its weighting is also low for a tech strategy, with only 60% of it in the US.

Winners have little time to run because it rebalances every six months, but it is important to note that equal weighting prevents it from becoming overly concentrated in tech megacaps. This helps diversify tech without exacerbating any overexposure to the largest players. However, excessive diversification can also be dangerous. Both winners and losers are inevitable, and with something this disruptive, the losers stand to lose greatly.

Active strategies, like those underlying the Allianz Technology Trust (LSE: ATT) or the Polar Capital Technology Trust (LSE: PCT), give seasoned managers the authority to decide which companies to invest in. Regarding the difficulties of investing in the upcoming AI phase, Dumas-Williams states, "You have to stay incredibly active, and frankly quite humble with your opinions." Being on the cutting edge requires you to stay extremely aware of developments.