But Boon is more than just an enthusiastic fan. As co-manager of the Neuberger Berman Next Generation Connectivity Fund, he counts Nvidia as the fund’s top equity position. The 1Q2026 fact sheet shows Nvidia accounted for 5.54% of the fund’s weight.
Few fund managers know semiconductors as intimately as Boon, a managing director and head of thematic, Asia, at Neuberger. The Hong Kong-based portfolio manager graduated with a degree in electronics engineering and started his career as a chip designer.
After stints at Arm, Cadence and Broadcom in the UK, he moved back to Asia to work at Taiwan Semiconductor Manufacturing Company-affiliated Global Unichip Corporation and Huawei subsidiary HiSilicon.
“I don’t look at things like a traditional manager,” Boon says. “I have a slightly different lens, not just on the investment or how we should look at a single stock but also the entire workflow, how we think, how we approach, how we look at the value chain, and how we do things.”
See also: The asset class hiding in plain sight
“I was an engineer by training. We didn’t have an IT department. We were the IT department. We don’t use Windows. We use the Unix system. We built our own systems,” he continues.
Recognising and grappling with these legacy issues has allowed Boon to take a more innovative approach to his work. He has relied more on his AI agents than his team over the past 12 to 18 months, particularly at weekends.
“Guess what? On the weekends, none of my analysts will respond to me. AI? No problem,” adds Boon. “Some people will say, ‘Oh, your AI could be wrong.’ Research analysts are always wrong, too. I used to be a research analyst; I know.”
See also: ‘Mission critical’ real estate makes the difference for investors: Pictet’s Gabert
“Being accurate is one thing, but knowing what’s going on, knowing how to make changes, what needs to change, what has changed in the industry — you have to be ahead of it. When we look at investments, it’s not just about what’s next. You should be a few steps ahead. What’s next after the next? That’s the key question I ask my whole team every day, not just what’s next. By the time you figure it out, you’re already too late.”
At a time when AI is driving the global economy, Boon has emerged as a prominent voice in the AI-driven cycle. Companies like Anthropic and OpenAI are targeting trillion-dollar IPOs. Memory giants Micron, Samsung Electronics and SK Hynix are also benefiting, with all three seeing strong valuation gains in May.
The surge of capital into AI has left investors questioning where the biggest opportunities lie. Boon says they should avoid focusing on individual companies and instead examine the broader AI value chain.
That evolution has been rapid. According to Boon, the AI industry has gone through several shifts in the past two to three years. In 2022 and 2023, the biggest winners were hardware manufacturers and server makers, which benefited from surging demand from companies such as Nvidia and saw sharp earnings upgrades. As valuations became stretched, investor attention shifted toward component suppliers and chipmakers.
“We’re now in the stage where good AI companies are those who have pricing power, and where there is a bottleneck in that specific industry,” he adds. “For example, memory, whether it’s DRAM (Dynamic Random-Access Memory) or NAND, falls into that area, but it’s not just that. Now we have many other components.”
Investors initially focused on top-tier companies in AI hardware and semiconductors. Attention has since shifted toward smaller firms, as leading manufacturers have approached the limits of their production capacity.
Boon believes major chipmakers in South Korea and Japan would welcome competition from lower-tier Chinese rivals. That would allow established players to concentrate on high-end data centre chips, while Chinese competitors meet demand for mid- to lower-end semiconductors used in smartphones and laptops.
For more stories about where money flows, click here for Capital Section
“If you look at a lot of the performances of these lower-tier suppliers, you’ll find that their share prices have done a lot better than the key leaders,” he says. “Even in the memory space, who is rallying more? Who is seeing that stronger earnings momentum? It’s actually the tier two and tier three players.”
AI race runs through Taiwan
The transformative and disruptive potential of AI has seen it morph into a national security issue for governments around the world. On June 12, Anthropic announced that it had been ordered by the US government to suspend access to its newly released Fable 5 and Mythos 5 models to foreign nationals both inside and outside the US. As a result, Anthropic disabled access to the models for all its customers.
China has taken similar measures as well. On April 27, Chinese regulators ordered AI start-up Manus to unwind its US$2 billion ($2.58 billion) acquisition by social media giant Meta. The company’s co-founders were reportedly barred from leaving China after they were summoned to Beijing for a meeting with China’s National Development and Reform Commission in March.
Boon argues that the gap between the US and China in AI isn’t so much a battle between the two countries. This is because most of the manufacturing expertise and capabilities are located in what he calls the “Golden Triangle of Asia,” which comprises Taiwan Japan, and South Korea.
“In terms of chip manufacturing, the US is actually far behind. The Chinese are also far behind,” he adds. “If you compare just chips manufactured on American soil versus Chinese soil, it’s almost the same. That’s why Taiwan is so important to both of them.”
Everybody’s asset-heavy
Software companies are often seen as highly attractive businesses to invest in because of their asset-light nature. After all, the marginal cost of shipping more copies of a piece of software is way lower than selling hardware products. But things appear to have flipped in the age of AI.
Tech giant Alphabet, Google’s parent company, has been ramping up capital expenditure in a bid to gain an edge in the AI race. Alphabet announced on June 2 that it had raised US$84.75 billion in equity capital to fund its AI investments. The deal was the largest secondary offering in history.
Boon says the shift by tech companies toward an asset-heavy structure is not new. It has been underway over the past decade with the rise of cloud technology services. “A lot of the workloads have shifted to the cloud. That’s when the internet giants have kind of stepped up a lot on their capex, and that has never come down,” he adds. “If you think about these internet companies, their competitive moat now is really the capex that they’ve put in.”
“Why is everyone looking to buy chips or chip services from Google, not just Nvidia? Because Google has been designing their own chips for the last 10 years. Their TPUs (Tensor Processing Units) are now so good. Had Google been asset-light and not focused on their TPU chips, they would not be where they are right now.”
Boon says retail investors are now looking ahead to a wave of AI IPOs this year, with many hoping to back the next Microsoft or Google of the AI era. He warns against this approach. “When it comes to investing, I don’t think it’s just about focusing on who would be the ultimate winner five to 10 years from now,” Boon adds. “For example, the internet giants, the Googles and the Amazons of the world, over the last 10 years, have evolved so much.”
Cheap doesn’t mean efficient
Initially, AI was seen as a way for companies to cut costs by hiring fewer workers. That view has weakened as businesses have confronted the high costs of using AI models in place of human employees.
Uber’s chief technology officer, Praveen Neppalli Naga, told technology news outlet The Information in April that the company had used up its entire annual AI budget in four months. According to a Bloomberg story published on June 3, the ride-sharing giant has since instituted an employee monthly usage cap of US$1,500 per AI tool.
Cost pressures are emerging across other companies as well, raising potential demand for lower-cost Chinese AI models. These models are cheaper on a per-token basis, partly because Chinese developers have had access to less compute and fewer US-designed chips, forcing them to build systems that are more resource efficient.
Chinese models may be cheaper than their American counterparts, but they are not necessarily more efficient. Tasks that Western models can complete in hours may take days on Chinese systems.
“The Chinese model could be cheaper, but it’s cheap for another reason,” he adds. “It’s cheap for the same reason, your Xiaomi phone and your iPhone are just of a different quality. Yes. It’s cheaper to get, but the output is going to be different.”
“More importantly, if you try to run these Chinese models, you will find that the quality is lower, not just because of the language model itself, but because there is insufficient computing capacity,” he adds. “You can’t wait for the model to come up with the code, say three days later, where the human being could have done it.”
Companies will ultimately have to introduce governance frameworks for AI use. Like electricity or water, AI will need to be used prudently rather than wastefully.
“If you want to waste electricity and water, you can do it. Obviously, the language model companies are going to do something to prevent this type of inefficient usage, because it burns them, too. So, there will be some guardrails down the line.”
