● Nvidia crushes AI bubble fears, supply chain power and profits emerge as real risks
Nvidia Rebutted the AI Bubble Narrative, but the Real Risks Are Supply Chain Constraints, Power, and Profitability
The key takeaway from this New York briefing is not simply that Nvidia reported strong earnings.
With robust results and a forecast for roughly 70% revenue growth, Nvidia directly addressed concerns about an AI bubble.
However, a close reading of the conference call suggests that the real issues for U.S. equities are not AI demand, but supply capacity, HBM pricing, data center power, and large-scale investment and guarantee exposure.
Bitcoin’s outlook is also increasingly tied to a broader macro narrative involving government debt and currency debasement.
Ahead of Jackson Hole, several Federal Reserve officials have also kept the door open to additional rate increases, adding caution to the Nasdaq and the AI semiconductor rally.
1. New York Market Summary: Nvidia Lifted the Nasdaq, but the Broad Market Did Not Rally
In premarket trading as of August 27, U.S. equity futures were led by technology stocks.
Nasdaq 100 futures rose about 0.87%.
S&P 500 futures gained about 0.32%.
Dow Jones futures fell about 0.19%.
This pattern suggests a rotation into AI-related equities rather than a broad-based risk-on move driven by macro optimism.
In other words, the advance reflected strength in Nvidia and the AI infrastructure chain rather than an across-the-board rally in U.S. equities.
Nvidia’s indication that fiscal 2028 revenue could rise by roughly 70% helped ease recent concerns about an AI investment slowdown.
At the same time, defensive sectors such as healthcare, financials, and consumer staples underperformed, underscoring the uneven nature of the move.
2. Key Asset Prices: Oil, Bitcoin, Gold, and the U.S. Dollar
In the Treasury market, prices for 30-year and 10-year bonds rose modestly.
Higher bond prices generally imply downward pressure on yields.
However, lower-than-expected weekly jobless claims signaled that the labor market remains resilient, which may limit the scope for lower yields.
European equities were mixed.
Germany’s DAX rose about 0.24%, while the Euro Stoxx declined about 0.39%.
Crude oil recovered modestly.
WTI rose about 0.33% to the mid-$85 range.
Brent increased about 0.81% to the high-$88 range.
Ongoing diplomatic efforts to ease tensions around the Strait of Hormuz have not fully removed supply disruption concerns.
Bitcoin rose about 1.02% to the high-$79,000 range.
Recent Bitcoin trading has been supported by concerns over U.S. fiscal deficits, rising government debt, and currency depreciation.
Gold traded broadly unchanged around $4,639 per ounce, and the dollar index was little changed at 99.12.
3. Sector Performance: Semiconductors, Memory, and AI Software Outperformed, While Apple, Financials, and Healthcare Weakened
Nvidia was the strongest performer by far.
In premarket trading, Nvidia rose about 6.63%.
Its long-term revenue outlook triggered buying across the AI semiconductor group.
Micron gained about 3.64%.
SK Hynix rose about 4.16%.
TSMC advanced about 1.27%, Broadcom about 1.3%, and AMD about 0.8%.
Memory and storage names also strengthened.
Seagate, SanDisk, and Western Digital all moved higher on expectations tied to AI data center demand.
Separately, SanDisk and Kioxia announced plans to invest $31 billion to expand production in Japan to meet AI memory demand, which further supported sentiment in the semiconductor space.
In software, Salesforce advanced on news of expanded collaboration with Anthropic.
Cybersecurity company CrowdStrike also rose after issuing favorable annual revenue guidance.
By contrast, Apple fell about 1%.
Microsoft also traded lower.
Among financials, JPMorgan declined, while healthcare and consumer staples were broadly weak.
Recent trading has shown a recurring pattern in which semiconductors outperform while defensive sectors lag.
4. U.S. Labor Market: Weekly Jobless Claims Came in Below Expectations
Initial jobless claims for last week were reported at 203,000.
That was below the market expectation of 208,000.
The decline in claims indicates that layoffs remain limited.
This supports the view that the U.S. labor market is still relatively firm.
That is constructive for growth.
However, from the Federal Reserve’s perspective, it reduces the case for near-term rate cuts.
If employment remains stable and inflation stays elevated, there is less incentive to ease policy quickly.
5. Nvidia Earnings: The Numbers Were Difficult to Fault
Nvidia reported second-quarter revenue of $96.2 billion.
That exceeded the market consensus of $91.9 billion.
Adjusted earnings per share came in at $2.22, also above expectations.
Data center revenue increased 18% quarter over quarter.
Roughly 93% of total revenue was generated by the data center segment.
Nvidia is increasingly operating as an AI data center infrastructure company.
Third-quarter revenue guidance was set at $108.0 billion.
This points to a strong likelihood that quarterly revenue will move above the $100 billion threshold for the first time.
The more important point is the outlook for next year.
Nvidia said revenue could increase by roughly 70% next year.
The key issue is that this figure reflects growth based on current supply capacity rather than unconstrained demand.
6. Nvidia Conference Call Q&A: Jensen Huang’s Response to the AI Bubble Debate
During the conference call, analysts focused less on the reported numbers and more on the durability of AI demand, supply bottlenecks, proprietary chip competition, profitability, and capital exposure.
These were the main takeaways from the earnings call.
6-1. Why Did Nvidia Provide a One-Year-Ahead Growth Outlook?
Morgan Stanley asked why Nvidia disclosed a one-year-ahead growth rate more explicitly than usual.
Jensen Huang said the issue is supply, not demand.
He noted that customer demand, if measured only by what clients want to buy, could be nearly double next year’s current level.
However, with the production capacity currently secured, the company believes it can confidently deliver about 70% growth.
In other words, 70% is not Nvidia’s view of the ceiling; it is a conservative estimate based on supply availability.
If supply expands further, actual growth could exceed that level.
6-2. How Much Will AI Agents Increase GPU Demand?
AI agents differ from conventional chatbots.
Instead of answering a single question, they can plan, search, use tools, and verify results autonomously once given a goal.
Jensen Huang said AI agents may require up to 100 times more compute than standard AI usage.
This means GPU demand can rise materially even without a proportional surge in the number of users.
The expansion of AI agents therefore remains one of the most important long-term drivers of AI semiconductor demand.
6-3. Would Growth Be Higher Than 70% If Supply Were Not Constrained?
Bernstein asked how high growth could be if supply were fully available.
Jensen Huang did not provide a specific number.
He said only that growth would be significantly higher.
He added that, based on customer requests, next year’s demand could rise to nearly twice current levels.
The formal guidance, however, is anchored in what Nvidia can actually supply.
6-4. Is Demand Concentrated Only Among Big Tech Customers?
The market often views Nvidia demand as concentrated among hyperscalers such as Amazon, Microsoft, Google, and Meta.
Jensen Huang said hyperscalers account for only about half of the picture.
The other half comes from enterprises, AI startups, and governments.
These customers typically cannot design their own chips or build fully integrated AI data centers.
As a result, they purchase not only GPUs, but also servers, CPUs, networking, and software as part of Nvidia’s full-stack platform.
This segment was described as generating quarterly revenue of about $40 billion, up 138% year over year.
That suggests Nvidia’s growth base is broadening beyond a narrow set of large cloud providers.
6-5. Is OpenAI’s Custom Chip a Threat to Nvidia?
Bank of America asked whether Nvidia risks financing a future competitor through its investment in OpenAI.
Jensen Huang said custom chips and Nvidia systems serve different purposes.
Custom chips are designed for specific inference tasks within a particular cloud environment.
Nvidia’s systems cover the full workflow, including data preparation, training, fine-tuning, inference, and production deployment.
They are also designed to operate across multiple clouds and data center environments.
The conclusion is that custom chips may replace some demand, but they are unlikely to displace Nvidia’s broader AI infrastructure platform.
6-6. How Large Are the Company’s Investment and Guarantee Commitments?
Analysts asked whether Nvidia’s investments, purchase commitments, and credit support could amount to as much as $500 billion over the next several years.
They also asked how much of that would translate into direct cash exposure.
The CFO said a significant portion of the commitments is concentrated in the first three years and is tied to next-generation production and supply chain buildout.
However, the company did not specify the total amount or the exact cash burden on Nvidia.
This remains an important risk factor for investors.
The extent to which Nvidia is financing AI partners, supporting data center construction, and indirectly enabling future product purchases is not yet fully transparent.
6-7. Would Open-Source AI Weigh on Nvidia?
UBS asked whether the growth of open-source AI could weaken closed, proprietary models such as OpenAI’s and ultimately reduce demand for Nvidia.
Jensen Huang responded that Nvidia benefits regardless of which model type prevails.
Closed models still run on Nvidia GPUs.
Open-source models used by enterprises and governments are also typically developed and deployed on Nvidia systems.
For Nvidia, the critical variable is not which AI model wins, but whether total AI usage continues to expand.
6-8. Could AGI and Self-Improving AI Further Increase Demand?
Analysts also asked whether self-improving AI and AGI-level systems could accelerate GPU demand.
Jensen Huang said demand growth could become faster.
Today, AI still responds to human prompts.
In the future, many AI agents may operate continuously in the background without direct human intervention.
He suggested that Nvidia, with about 40,000 employees, could eventually rely on 400,000 or even 4 million AI agents.
That would imply a substantial increase in compute demand without a corresponding increase in headcount.
6-9. Three Factors More Important Than the Exact Timing of AGI
Jensen Huang said some tasks already appear close to AGI-like capabilities.
However, he argued that pinpointing the exact date of full AGI is not especially meaningful.
Instead, he highlighted three questions:
First, is AI doing useful work?
Second, are AI service providers generating revenue?
Third, does adding more GPU capacity increase revenue and profit?
He argued that these conditions are already being met.
In his view, AI investment is not driven only by distant expectations, but by current commercial bottlenecks that can be relieved with additional compute.
6-10. What Is the Main Constraint: GPUs, Memory, or Power?
One of the most important questions concerned the source of supply constraints.
Analysts asked whether the bottleneck lies in semiconductor manufacturing, HBM memory, site availability, power, cooling, water, or networking.
Jensen Huang declined to identify a single constraint and said the entire supply chain is effectively operating at capacity.
More GPUs do not help if there is insufficient HBM.
Servers cannot be deployed if there is not enough power, cooling, or data center space.
This suggests Nvidia’s main risk is not a collapse in demand, but the inability of the ecosystem to keep pace with orders.
In the United States, local resistance to data center construction, pressure on the power grid, and political concerns are also making this issue more complex.
6-11. Revenue Potential per 1 GW Data Center Continues to Rise
Another important point was revenue potential per 1 GW of data center power.
Under the Hopper generation, Nvidia could generate about $18 billion of revenue per 1 GW data center.
With Grace Blackwell, that figure rises to about $25 billion.
With the Vera Rubin generation, it could reach about $40 billion.
This reflects the expanding scope of Nvidia’s product stack over time.
The company is moving beyond GPUs into a broader AI data center platform spanning CPUs, networking, servers, and software.
7. A Margin Headwind: Profitability Could Come Under Pressure
Nvidia’s current gross margin is around 75%.
However, rising HBM memory prices could push that figure down to roughly 71% to 72% in the fourth quarter.
Revenue is growing rapidly, but the cost of critical components is also rising.
That means margin preservation is becoming increasingly important.
Investors should focus on three issues going forward:
First, how quickly Nvidia can expand supply capacity.
Second, how effectively it can pass higher HBM costs through to customers.
Third, how much capital strain may come from AI-related investments and guarantees.
8. Dollar Tree and Dollar General: Similar Discounters, Divergent Stock Performance
Premarket attention also centered on U.S. discount retailers Dollar Tree and Dollar General.
Both sell low-priced household goods and general merchandise in a dollar-store format.
Dollar Tree fell about 3.4%.
Dollar General rose more than 4%.
Both companies reported solid results, but tariff-related reimbursement effects drove the divergence.
Refunds linked to the Trump administration’s tariff regime were reflected in earnings.
Dollar Tree reported EPS of $2.70, but about $1.31 of that came from tariff reimbursements.
Because the benefit was largely non-recurring, the market viewed the quality of earnings as weaker.
Its third-quarter EPS outlook was also below expectations.
By contrast, Dollar General saw only about $0.25 per share from tariff reimbursements.
Its underlying operating performance was stronger even after excluding the one-time benefit, and it raised its full-year earnings outlook.
As a result, investors responded more positively to Dollar General.
9. The Case for a $500,000 Bitcoin: Government Debt and Currency Debasement
Bernstein said Bitcoin could reach $150,000 by mid-next year, and $200,000 to $300,000 in a bull market.
For 2029, the firm’s base target is $300,000, with a bullish case of up to $500,000.
The central argument is rising government debt.
Over the past several decades, falling interest rates made it easier for governments to add leverage.
But higher rates increase debt-servicing costs.
Governments can reduce debt only by raising taxes or cutting spending.
Politically, that is difficult.
As a result, they may be incentivized to allow currency debasement through easier monetary policy or higher inflation.
This is often described as the debasement trade.
It refers to buying scarce assets such as gold or Bitcoin as a hedge against currency dilution.
Bernstein also pointed to the resilience of long-term Bitcoin holders, noting that many did not sell even during large drawdowns.
It also cited potential institutional inflows through spot ETFs as a support factor.
That said, a $500,000 Bitcoin scenario by 2029 is highly aggressive.
It is better understood as a macro thesis about debt accumulation and currency debasement than as a near-term price forecast.
10. Jackson Hole: The Market’s Focus Has Shifted to the Federal Reserve
The Jackson Hole Economic Policy Symposium has begun.
It brings together central bankers and policymakers from around the world to discuss the direction of monetary policy.
This year’s theme is the impact of financial innovation on payments and policy.
The key event will be the speech by Kevin Warsh on August 28 at 10:00 a.m. Eastern Time, or 11:00 p.m. Korea time.
Markets are focused on how seriously he views inflation and whether he believes current rates are sufficiently restrictive.
11. Hawkish Warnings Ahead of Jackson Hole: Rate Hikes Were Not Ruled Out
Kansas City Fed President Jeff Schmid said inflation remains too high and that current rates may not be restrictive enough.
He even suggested that policy could still be more accommodative than intended.
He cited oil prices, data center construction, raw material demand, and strong domestic demand as upward pressures on inflation.
Core PCE remains well above the Federal Reserve’s 2% target, which is also a concern.
Schmid did not fully rule out the possibility of a 25 basis point or even 50 basis point rate increase.
However, he does not have a vote this year and has long been among the more hawkish voices on the Committee.
Accordingly, his remarks are best read as evidence that inflation concerns remain elevated inside the Fed rather than as a new policy signal.
12. Key Takeaways Often Overlooked in the News Flow
First, Nvidia’s 70% growth outlook reflects supply constraints more than demand limits.
Most headlines focus on the revenue figure itself, but the more important interpretation is that customer demand could be even stronger.
Second, the real risk for Nvidia is not an AI bubble collapse but bottlenecks in power and data center buildout.
More GPUs do not translate into more revenue if HBM, cooling, land, grid capacity, or permitting become binding constraints.
Third, AI agents are a key variable because they expand compute usage faster than user count alone would suggest.
Even if the number of AI users does not rise sharply, 24/7 background workloads can materially increase GPU demand.
Fourth, custom chip competition is unlikely to replace Nvidia entirely.
Even if OpenAI, Google, Amazon, or Microsoft build their own chips, those designs are generally narrow in scope, while Nvidia supplies the broader AI data center platform.
Fifth, the bullish Bitcoin thesis is not just speculative; it is tied to fiscal deficits and currency debasement risk.
The more important point is that the debasement trade is reemerging as a major market theme.
Sixth, after Jackson Hole, markets may reassess the balance between the AI rally and rate pressure.
Even if Nvidia continues to support the Nasdaq, a reversal in rate-cut expectations could weigh on U.S. equities more broadly.
< Summary >
Nvidia beat expectations on revenue, earnings, and guidance.
The AI bubble narrative weakened in the near term, but supply chain constraints, HBM pricing, and data center power remain the main risks.
AI agents could require up to 100 times more compute than conventional AI, supporting long-term demand for AI semiconductors.
Bitcoin’s outlook is being framed around government debt growth and currency debasement.
Ahead of Jackson Hole, hawkish Fed commentary has brought the policy path back to the forefront for U.S. equities.
[Related Articles…]
- Nvidia and the AI Semiconductor Supercycle: Data Center Investment Outlook
- Bitcoin Outlook and the Debasement Trade in a Weak-Dollar Era
*Source: [ Maeil Business Newspaper ]
– 엔비디아, AI 거품론에 답했다ㅣ컨콜 Q&A 분석ㅣ비트코인 50만달러론, 왜?ㅣ고용시장 여전히 견조ㅣ잭슨홀 전야 매파 경고ㅣ홍혜진의 뉴욕브리핑
