By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
Panelists generally agree that NVDA's $350 target by 2028 is plausible but risky, hinging on assumptions about AI demand, hyperscaler capex, and margin compression from custom silicon.
Risk: Margin compression from hyperscalers' custom ASICs and potential normalization of hyperscaler capex.
Opportunity: Potential extension of the capex super-cycle due to sovereign AI buildouts.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
If it were easy to predict stock prices, there would be far more rich people out there. Fortunately, investors can make great money by simply gauging where a stock might go in the future, based on the underlying company's growth opportunities and valuation. If you're right about the factors moving a stock, you'll probably do well, even if you don't nail the exact share price.
For example, I predict that Nvidia (NASDAQ: NVDA) will trade at $350 per share by 2028. Keep in mind that Nvidia's fiscal calendar ends in January. Therefore, its fiscal year 2028 ends just as the calendar year 2028 begins.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue »
Remember, the reasons why I believe Nvidia can soar 71% higher over the next 18 months are far more important than whether the stock trades at $350, $325, or $450. As long as my prediction about Nvidia is directionally correct, investors will be happy they bought the stock. I'll unpack how I arrived at my prediction below.
Nvidia has generated $253.5 billion in trailing 12-month revenue and continues to pump out breathtaking growth. The company's sales grew by 85% year over year in the first quarter of Nvidia's fiscal year 2027, driven by Grace Blackwell, its current flagship data center chip platform. Its successor, Vera Rubin, has entered full production and could begin shipping later this year.
CEO Jensen Huang has said that Nvidia anticipates $1 trillion in sales between Blackwell and Rubin through next year, a clear signal that artificial intelligence (AI) hyperscalers haven't relented from pouring billions of dollars into AI compute. Based on Wall Street estimates, Nvidia's total sales could more than double by fiscal year 2028, to approximately $555.5 billion.
Analysts also estimate that Nvidia will earn $12.79 in fiscal year 2028, nearly double Nvidia's trailing 12-month earnings per share of $6.53. That seems reasonable as long as the company maintains its pricing power. A valuation of about 27 times those earnings produces that $350 share price. Keep in mind that those will be trailing 12-month earnings by January 2028. Today, Nvidia trades at 31 times its trailing earnings, so this prediction assumes its valuation will decline.
The numbers can change. Rubin consists of seven chips, forming an AI supercomputer chipset that expands Nvidia's data center footprint. There could be even more growth and upside if Rubin exceeds sales expectations. On the flip side, hyperscalers could pull back on their capital expenditures, ending this data center boom at any moment. The uncertainty is simply part of the game.
Four leading AI models discuss this article
"Nvidia's explosive growth trajectory faces material concentration and cyclical risks that the article underplays, so a 27x multiple on doubled earnings is not a foregone conclusion."
The article's $350 NVDA target by early 2028 implies only ~71% upside from current levels (~$205 post-split) while forecasting revenue to more than double to $555B and EPS to ~$12.79. That requires the multiple to compress from today's 31x trailing to 27x, which is plausible but far from assured. The piece correctly flags AI capex momentum via Blackwell/Rubin yet glosses over concentration risk: data-center revenue is ~88% of total, hyperscalers (MSFT, AMZN, GOOGL, META) drive the bulk, and any slowdown in their AI ROI could truncate the cycle well before FY28. Missing context: historical GPU boom-bust patterns and potential sovereign or open-source competition eroding Nvidia's pricing power.
The strongest case against this modest-upside call is that if Rubin drives even 10-15% above consensus and gross margins hold above 75%, the stock could re-rate to 35-40x forward earnings, pushing shares well past $450 by 2028—making the $350 target look conservative rather than realistic.
"Nvidia's valuation is currently tethered to an unsustainable pace of hyperscaler capex that ignores the inevitable shift toward custom silicon and the need for demonstrable AI ROI."
The article’s $350 target for NVDA by 2028 is mathematically plausible but strategically naive. While a 27x forward P/E (price-to-earnings ratio) is a conservative multiple for a high-growth leader, the analysis ignores the inevitable margin compression that occurs as hyperscalers like Microsoft and Google develop custom silicon (ASICs) to reduce dependency on Nvidia. Furthermore, the $555 billion revenue projection assumes a linear continuation of current AI infrastructure spending, failing to account for the 'utility phase' where companies must prove ROI on AI before committing to the next massive capex cycle. Nvidia is priced for perfection, and any deceleration in data center demand will trigger a violent multiple contraction.
If Nvidia successfully transitions from a hardware vendor to a recurring software and services platform, the current valuation might actually be too low, as they would capture a larger share of the AI value chain beyond just the chip sale.
"The $350 price is defensible only if Nvidia's gross margin holds above 70% and capex cycles don't normalize—neither is guaranteed and both are underexplored in this analysis."
The $350 target rests on two fragile assumptions: (1) Nvidia maintains 27x forward P/E despite maturation, and (2) revenue doubles to $555B by FY2028. The article acknowledges valuation compression from today's 31x but doesn't stress-test *why* that multiple contracts—typically it's margin pressure, competition, or saturation. The $1T Blackwell/Rubin revenue claim is CEO guidance, not booked. Vera Rubin's 'full production later this year' is vague; execution risk on a 7-chip supercomputer is real. Most critically: the article assumes hyperscaler capex doesn't normalize. If it does—and history suggests it will—the entire revenue thesis collapses.
If hyperscalers face ROI pressure or AI monetization stalls, capex cycles compress sharply (see 2016-2017 datacenter slowdown). Nvidia's gross margins could compress 300-500bps if competition intensifies or customers shift to custom silicon, making $12.79 EPS by FY2028 optimistic.
"Nvidia can deliver higher EPS through 2028, but the stock's upside hinges on a resilient AI capex cycle and a forgiving multiple; without it, the $350 target is at risk."
Today's piece is bullish: it leans on Nvidia's data-center turbocharged growth and a forward-looking EPS of $12.79 by FY2028 to justify a $350 target, implying a ~27x multiple on forward earnings. It hinges on durable AI-data center demand from hyperscalers and successful ramp of Blackwell/Rubin chips. But the missing context is critical: AI capex is highly cyclical, and a turn in macro demand or a policy clamp on AI hardware (e.g., export controls to China) could throttle orders. Competitive risk from AMD/Intel GPUs, margin pressure from supply chain costs, and a regime of multiple compression if rates rise also threaten the thesis.
AI capex could peak earlier or turn down, especially if policy/export controls or a macro downturn hit hyperscalers. Any deterioration in Nvidia's data-center market share or in its margin due to supply costs could trigger substantial multiple compression, making $350 unlikely even with solid EPS growth.
"Sovereign AI demand could extend Nvidia's cycle beyond hyperscaler ROI constraints."
Gemini's margin compression from custom ASICs is real but timeline-misplaced: hyperscalers' in-house silicon won't scale to displace 80%+ of Nvidia's CUDA moat until 2029+. The real unmentioned risk is sovereign AI buildouts (Saudi, UAE, China) creating parallel demand that bypasses hyperscaler ROI scrutiny entirely, potentially extending the capex super-cycle past 2028.
"Geopolitical export controls and systemic manufacturing complexity represent greater threats to Nvidia's margin stability than the competitive threat of custom ASICs."
Grok, your sovereign AI thesis is compelling but ignores the geopolitical 'kill switch.' If the US government views sovereign clusters in the Middle East as national security threats, export controls will tighten, effectively neutralizing that demand buffer. Meanwhile, Claude correctly identifies the 'execution risk' of the Rubin supercomputer. If the architecture shifts from chip-level to system-level, Nvidia’s supply chain complexity increases exponentially, making the 75%+ gross margin target a massive, unpriced liability.
"Export controls won't kill sovereign demand in time to prevent margin compression from utilization risk in a mid-cycle demand dip."
Gemini's export-control 'kill switch' is real, but the timing assumption is backwards. US policy typically *lags* geopolitical risk by 18-24 months. Sovereign clusters in Saudi/UAE are already operational; controls would come after capex peaks, not before. This actually *extends* the cycle Gemini fears will compress. More pressing: nobody flagged that Nvidia's gross margins depend on utilization rates. If Blackwell ramps but demand softens mid-2026, fixed costs spike and margins crater faster than EPS grows—the $12.79 target becomes unreachable even if revenue hits $555B.
"The main risk to the $350 target is a mid-cycle demand slowdown that compresses multiples before margins deteriorate."
Gemini's concern about margin compression from custom AI chips is valid, but the bigger flaw is timing. Even with hyperscaler capex normalization, Nvidia's CUDA moat and software monetization can cushion margins, and a decline in hardware demand may hit multiples before EPS—not after. The underappreciated risk is a mid-cycle demand slowdown triggering a sharp multiple re-rating, even if 2026-28 EPS grows, which could render the $350 target too optimistic.
Panelists generally agree that NVDA's $350 target by 2028 is plausible but risky, hinging on assumptions about AI demand, hyperscaler capex, and margin compression from custom silicon.
Potential extension of the capex super-cycle due to sovereign AI buildouts.
Margin compression from hyperscalers' custom ASICs and potential normalization of hyperscaler capex.