Renowned economist and Chief Investment Officer at Lotus Asset Management, Hong Hao, shared his latest market outlook during a Now Finance interview on September 1, addressing topics ranging from AI industry returns and competitive dynamics to upstream resource demand. In a subsequent etnet discussion on August 27, he also delved into equity market performance, the U.S. Treasury's long-dated bond buybacks, and the trajectory of gold prices.

Hong noted that the strong rebound in Hong Kong stocks has pushed the Hang Seng Index above the 25,000 mark, though it has encountered some temporary resistance at these levels. He remains confident that the third and fourth quarters will bring forth a greater abundance of investment opportunities.

Regarding whether the massive capital expenditures by AI giants can yield adequate returns, Hong believes the market has grown accustomed to the narrative of first "pouring in capital and seizing territory" before evaluating profitability. Given the widespread conviction that AI demand is immense, investors assume future returns will eventually materialize even after such substantial investments. However, he cautioned that with annual AI investments ranging between $1.2 trillion and $1.5 trillion and a depreciation period of eight years, the seven major tech companies would collectively need to generate $8 to $10 trillion in annual revenue to justify these outlays. This translates to every one of the world's 8 billion people spending at least $1,000 each year on AI services, a threshold that seems improbable when considering many regions where annual incomes barely reach $100.

Addressing the hot themes of price increases and supply shortages in certain semiconductor segments, Hong warned that emerging technologies may circumvent these bottlenecks. For instance, NVIDIA's liquid-cooled servers reportedly require significantly less High Bandwidth Memory (HBM), reducing the layer count from 12 to 8, with speculation of even further reductions. Alternatives such as DRAM or NAND flash memory could potentially replace HBM entirely. Hong emphasized that technological breakthroughs have historically overcome supply constraints, and this cycle is unlikely to prove exceptional.

AI Industry Narrative: Capital Deployment Before Returns

When asked whether investors have adjusted their expectations for when substantial AI investments will translate into profitability, Hong observed that the market has become accustomed to a playbook reminiscent of China's infrastructure and property boom — securing land first and calculating profits later. Since everyone believes AI demand is extraordinarily strong, current investments are viewed as stepping stones to future returns. Yet even the largest technology companies are presently witnessing robust growth accompanied by compressed profit margins.

Innovative Technologies May Bypass Semiconductor Constraints

Hong elaborated that while the prevailing view holds that strong AI demand has created semiconductor chip shortages, new technologies could potentially sidestep these limitations. Beyond NVIDIA's adjustments to HBM requirements, other innovations like DRAM or NAND alternatives are emerging rapidly. If a major player reduces its appetite for premium HBM, competitors would likely follow suit, potentially undermining the assumption of sustained supply scarcity in semiconductors. History suggests technological progress tends to overcome supply deficits, and Hong believes this pattern will persist.

Upstream Commodities Also Poised for Growth

Despite the focus on AI's downstream applications, Hong highlighted that upstream sectors — particularly energy and base metals — may be underappreciated. Copper and aluminum have already reached multi-year or record highs this year, and speculative activity has extended into agricultural commodities due to ongoing conflicts and climate-related disruptions. If AI demand continues its exponential rise, humanity's appetite for commodities and metals should expand correspondingly. Without these raw materials — from copper to rare earths — semiconductor manufacturing becomes impossible. Thus, if one anticipates semiconductor supply shortages, upstream energy and commodity markets should face even tighter conditions.

Low Barriers to Entry for Large Language Models

Hong dismissed the notion that large language model companies possess meaningful competitive moats. Many firms deploy proprietary local models tailored to specific tasks, creating a fragmented competitive landscape. The frequent release of new models every few weeks demonstrates the absence of entry barriers. Furthermore, Chinese companies have achieved comparable performance at a fraction of Western costs, with token prices potentially 10 to 20 percent of Western levels. Consequently, most dedicated model providers currently struggle to generate profits.

Platform Companies to Emerge as Ultimate Winners

The most likely victors in the AI value chain, Hong posited, are existing platform companies. These firms control essential user traffic gateways — analogous to WeChat's Yuanbao or Douyin's Doubao — without having made colossal upfront investments. By purchasing ready-made models and reselling tokens at a margin, they enjoy structurally advantageous positions. The skepticism surrounding the seven tech giants stems from their transformation from cash-rich enterprises into negative cash flow entities after years of heavy spending without commensurate revenue visibility.

More Opportunities Await in Q3 and Q4

Hong reiterated that the Hong Kong market's rally above 25,000 has encountered minor headwinds, but he expects additional openings to surface during the latter half of the year. Investors who have yet to capitalize should remain patient, as the year is far from over — schools only just reopened on September 1, leaving several months for market performance to unfold.

Treasury Buybacks Fuel Dollar Weakness and Risk-On Sentiment

While bond yields may not have peaked independently, a significant development has emerged: the U.S. Treasury intervened in the bond market by repurchasing $2 billion in long-dated securities, with plans to double that amount to $4 billion. Older issues often lose liquidity over time as holders retain them until maturity or use them as collateral, rendering their pricing inefficient. By selectively buying these illiquid bonds, the Treasury can effectively influence market pricing with modest capital outlays. This maneuver carries considerable potency. If long-term yields decline, the dollar should weaken — a trend already manifesting — which would likely boost risk assets, particularly commodities, precious metals, gold, silver, and Bitcoin. Although such measures haven't been fully implemented yet, Hong expressed confidence in the Treasury's capacity to follow through.

Gold's Unique Dual Nature as Ultimate Safe Haven

Hong emphasized gold's distinctive characteristic of functioning as both a risk asset and a safe haven, unlike Bitcoin, which purely behaves as a risk asset. During ordinary periods, falling interest rates prompt gold to appreciate alongside risk assets such as Bitcoin. However, in times of heightened market stress or tail risks, gold serves as the ultimate hedge. Importantly, gold protects against extreme inflation rather than modest price increases — for instance, a sudden surge from 3 percent to 10 percent would trigger a bull run in gold, reminiscent of the 1970s oil crisis.

Long-Term Uptrend Intact; Pullbacks Create Entry Points

Hong reiterated his long-standing projection from over a year ago that a favorable buying opportunity for precious metals and cryptocurrencies would emerge around September or October this year. Gold has already appreciated significantly from $3,900 to over $4,700, with futures reaching approximately $4,740 on August 25. Statistically, such levels invite profit-taking and consolidation. Rather than questioning current opportunities, investors should prepare for potential pullbacks — if gold retreats to $4,400 after reaching $4,800, some may lose confidence. Yet a substantial correction would actually generate superior entry points, making such declines advantageous rather than concerning.