For the past year, the market has been fixated on the idea that NVIDIA’s meteoric rise, fuelled by its dominance in AI hardware, must inevitably end in a crash. The consensus has been that NVIDIA’s high margins, driven by its graphics processing units (GPUs) and AI accelerators, are unsustainable, and that competitors or market saturation will eventually erode its position.
Yet, as Bloomberg columnist John Authers astutely notes, the bubble isn’t in NVIDIA’s shares but in the insatiable demand for its chips. The market’s persistent expectation of a burst has paradoxically kept the bubble at bay: a bubble isn’t a bubble when everyone anticipates its collapse, nor can it burst when scepticism dominates sentiment.
This paradox is central to understanding NVIDIA’s current trajectory. The market’s reluctance to fully embrace NVIDIA’s business model has (so far) kept its stock price from reaching the euphoric levels that characterise true bubbles. Instead, NVIDIA’s valuation reflects a grudging acknowledgment of its dominance, tempered by fears of competitive threats and geopolitical risks.
However, recent developments suggest that the market may soon shift from scepticism to conviction, driven by NVIDIA’s unmatched technological moat, strategic foresight, and the explosive growth of AI demand.
Earnings Insights: A Glass Half Full
NVIDIA’s recent earnings may provide a compelling case for reevaluating the narrative. The market’s reaction to the company’s Q1 2025 numbers reveals a “glass half full” sentiment, where even setbacks are spun into positives. For instance, US restrictions on sales to China forced NVIDIA to write down its inventory of H20 chips by US$4.5 billion and forgo an estimated US$2.5 billion of sales while the company estimates the lost revenue opportunity for Q2 to be approximately $8 billion.
Unless there is a change of heart in the White House, this lost revenue and write-down are not coming back. There is at best a limited market for the H20, and possibly no market whatsoever, but this seems to be of no concern. Wall Street loves an apples-to-bananas comparison, and while guidance for revenue of US$45 billion is in line with expectations, the exciting ‘beat’ comes from adding back the US$8 billion of unavailable revenue. Whatever, the numbers are still impressive and there has been a clear acceleration.
This sentiment is further bolstered by NVIDIA’s financial health. The company generated US$26 billion in free cash flow (FCF) in the quarter, using US$14 billion to buy back stock – an unprecedented move at this scale. CEO Jensen Huang’s aggressive buyback signals confidence in NVIDIA’s long-term prospects and suggests he knows something the market does not.
Additionally, Huang addressed ongoing concerns about NVIDIA’s reliance on Singapore for 20% of its sales, clarifying that many large customers use Singapore for centralised invoicing while products are shipped globally. This transparency puts NVIDIA on the hook for clarity surrounding the final destination of its chips, which will be distracting.
The Infinite Game of AI
One of the most striking moments from NVIDIA’s earnings call was Huang’s direct appeal to the Trump administration. He urged policymakers not to cut China out of the AI ecosystem, emphasising that forcing half the world to source silicon elsewhere would harm US interests.
Huang framed AI as an “infinite game” in that, unlike traditional industries with clear winners and losers, AI is a continuous, evolving frontier with no definitive end. This perspective contrasts with Washington’s “finite game” mindset, which focuses on short-term victories and dominance. By positioning NVIDIA as a key player in this infinite game, Huang is further aligning the company with the long-term, global growth of AI.
This vision is backed by tangible demand. Huang disclosed that nearly 100 new AI factories are in development this quarter, doubling the number of GPUs in each cluster and expanding globally. This growth isn’t limited to China; it spans Europe, the Middle East, and beyond.
For example, Microsoft has already deployed tens of thousands of NVIDIA’s Blackwell GPUs and is expected to scale to hundreds of thousands, with Open AI remaining a key customer. NVIDIA’s CFO Colette Kress noted that hyperscalers are deploying 1,000 GB200 NV72 racks (72,000 GPUs) per week, a run rate that translates to four million GPUs annually. This scale underscores the relentless demand for NVIDIA’s products, driven not just by hyperscalers but by a broadening customer base.
Dismantling the Bear Case
The bear case against NVIDIA rests on three pillars: unsustainable margins, high customer concentration among hyperscalers, and the competitive threat from application-specific integrated circuits (ASICs). However, each of these arguments is losing steam considering NVIDIA’s performance and market dynamics.
First, the notion that NVIDIA’s “supernormal” margins cannot last is being challenged by reality. Despite the China-related write-down, NVIDIA’s margins remained stable and were guided higher, defying expectations of erosion.
Second, while NVIDIA does rely heavily on hyperscalers like Microsoft and Google, the rapid expansion of AI factories and new customer markets – such as the Middle East soaking up China’s slack – suggests that customer concentration is diluting.
Finally, the threat from ASICs (which allow for the creation of deeply differentiated silicon solutions for customers), particularly from partners like Broadcom (AVGO), is overstated. Google’s Tensor Processing Unit (TPU), the leading AI ASIC, struggles to gain traction with third-party customers, and even Google has signed a significant new contract with CoreWeave, a cloud provider heavily reliant on NVIDIA GPUs.
The broader AI landscape further undermines the bear case. At recent industry events like Microsoft Build and Google I/O, CEOs Satya Nadella and Sundar Pichai highlighted the explosive growth of AI inference. Google reported processing 480 trillion tokens monthly, a 50x increase year over year, while Microsoft processed 100 trillion tokens in a single quarter, up 5x from the previous year. This surge in AI usage underscores the growing demand for NVIDIA’s GPUs, which remain the gold standard for both training and inference workloads.
The Total Cost of Ownership Moat
NVIDIA’s most formidable weapon is its Total Cost of Ownership (TCO) advantage, particularly with its GB200 NVL72 system. Described as a TCO monster by analysts, the GB200 delivers massive performance gains and a 30% reduction in power and thermal load through its tightly integrated, liquid-cooled design. In an industry where power consumption is a critical bottleneck, especially in the US and other developed markets – this efficiency is a game-changer. Competitors attempting to replicate NVIDIA’s connectivity and coherency without its proprietary NVLink technology face significant power and cost penalties, making it nearly impossible to match NVIDIA’s TCO.
This advantage gives NVIDIA leverage over hyperscalers. Huang has strategically increased sales to secondary players like CoreWeave, perhaps signalling to hyperscalers that the company can prioritise other customers if pushed too hard on margins. Indeed, such is the demand for the GB200 that scarcity may be behind Google’s recent contract with CoreWeave, which appears to have received a significant number of the in-demand box.
Google can scarcely afford to hit bottlenecks in the face of a wall of demand, especially given traditional search is under pressure. By controlling the supply of cutting-edge hardware, NVIDIA ensures that hyperscalers remain dependent on its ecosystem, even as they explore alternatives.
The Path to a Real Bubble
Ironically, the conditions for a true NVIDIA bubble may only emerge once the market fully embraces its dominance. The current scepticism, coupled with the bear case’s weakening arguments, suggests that NVIDIA’s stock price has not yet reached the euphoric levels that precede a crash. Instead, as AI demand continues to explode and NVIDIA’s technical barriers, such as its TCO moat and rapid product development, prove insurmountable, the market may shift to a “hurrah” mentality, where conviction replaces doubt.
This shift could be catalysed by broader market appreciation of NVIDIA’s unassailable position. Competitors like ASICs face a steep climb to challenge NVIDIA’s ecosystem, and the company’s ability to deliver consistent innovation ensures that it remains ahead of the pack. As AI adoption accelerates and new markets emerge, NVIDIA’s customer base will diversify, further reducing risks tied to hyperscaler concentration.
If the market recognises that NVIDIA’s dominance is not a fleeting peak but a sustainable “eyrie”, the stage could be set for a bubble driven by herd conviction rather than speculative frenzy.
NVIDIA’s story is not one of an imminent collapse but of a company redefining the AI landscape. Its financial strength, strategic vision, and technological moat position it as a leader in an infinite game where demand shows no signs of slowing. The market’s current scepticism may soon give way to enthusiasm as NVIDIA’s TCO advantage, global expansion, and relentless execution become undeniable.
Far from a bubble waiting to burst, NVIDIA may be on the verge of cementing its place as an unassailable force in AI, with the potential for a true bubble only emerging once the market fully buys in. For now, NVIDIA’s legs are not just holding – it’s sprinting ahead.
Tim Chesterfield is the long-time CIO of the Perpetual Guardian Group and the founding CIO and Director of its investment management business, PG Investments. With $2.6 billion in funds under management and $8 billion in total assets under management, Perpetual Guardian Group is a leading financial services provider to New Zealanders.
Perpetual Guardian Group recently acquired boutique fund manager Castle Point, which now forms part of the Group’s investment management suite of businesses.
Disclaimer
Information provided in this publication is not personalised and does not take into account the particular financial situation, needs or goals of any person. Professional investment advice should be taken before making an investment. The information provided in this article is not a recommendation to buy, sell, or hold any of the companies mentioned. PG Investments is not responsible for, and expressly disclaims all liability for, damages of any kind arising out of use, reference to, or reliance on any information contained within this article, and no guarantee is given that the information provided in this article is correct, complete, and up to date.
This article was originally published by the NBR. You can read the original piece here.
