The compute node for the next era. The stock prices in flawless execution.
Revenue is up 85% year-over-year ($81.6B in Q1 alone) and free cash flow sits at an astronomical $50B for the quarter — yet NVDA has pulled back from its all-time highs as the market wrestles with a single question: are we building the permanent infrastructure layer of the 21st century, or standing at the precipice of a cyclical capex cliff? Four analyst lenses, three scenarios, four time horizons.
Gray line = NVDA's actual price architecture into today ($236 peak May '26 → $204.65 now); colored paths = synthesized scenario midpoints forward, probability-weighted (base 50% · bull 25% · bear 25%). Log-linear, mid-year marks. Wall Street 12-month consensus ≈ $274.
Re-weight the scenarios
The $5 Trillion question is how durable the Blackwell ramp is versus a potential capex hangover. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target below, the dotted line on the chart, and the prob-weighted row of the scenario cards all update live.
Four analyst lenses, four answers
The exact same silicon and financial statements support wildly different conclusions depending on which structural framework you trust. Each lens below is a synthesized expert perspective with its own 12-month target.
The Compute Supercycle
Hyperscaler capital-expenditure plans have eclipsed $725 billion. The Blackwell GB300 NVL72 ramp is the fastest in company history. We are not watching a hardware cycle; we are watching the buildout of a $3-4T annual AI infrastructure market by 2030. At 85% revenue growth, the current multiple is actually cheap given the compounding EPS profile.
The Cash Machine
Nvidia just printed $50.3B in operating cash flow in a single quarter with ~75% gross margins. The cash engine is so powerful they authorized an $80B buyback and hiked the dividend by 25x. Even if hyperscaler demand decelerates into mid-teens growth, the sheer volume of capital return and pristine balance sheet provides a massive fundamental floor.
The Cyclical Short
Trees do not grow to the sky. Roughly 50% of Data Center revenue is concentrated in a handful of hyperscalers whose AI revenue ROI isn't matching their GPU spend. Once those clusters are built, orders pause. Factor in zero Data Center revenue from China and rising custom ASIC competition (Google TPU, Meta MTIA), and a multiple compression is inevitable when growth turns negative.
The CUDA Monopoly
Bears miss that Nvidia isn't selling chips; they are selling entire "AI Factories." The competitive advantage isn't just the silicon architecture—it's the CUDA software ecosystem, the InfiniBand/Spectrum-X networking tie-ins, and the developer lock-in. Switching costs are astronomical. Sovereign AI and Enterprise adoption are diversifying them away from purely hyperscaler reliance.
Wall Street 12-month price targets
What the sell-side expects over the next year. Bars are sorted low to high; the dashed line is today's $204.65 — note how virtually every major desk views the current level as a heavy discount.
Sell-side 12-month targets — consensus sits at ≈ $274, representing about +34% upside. An overwhelming ~97% of analysts covering the stock rate it Buy/Strong Buy. The dashed line marks today's $204.65: the gap between the live price and the lowest estimates indicates market unease that modeling hasn't fully captured. Firms, ratings, and targets are illustrative compositions of current coverage.
Where the compute cycle leads
Synthesized scenario midpoints (mid-year). Returns shown vs. today's $204.65. These are illustrative frameworks based on the node architecture of earnings and exit multiples—five-year outcomes hinge strictly on whether hyperscaler capex represents a structural plateau or a cyclical peak.
1 Year
Mid-20272 Years
Mid-20283 Years
Mid-20295 Years
Mid-2031▸ Bull case — show the assumptions & math
▸ Base case — show the assumptions & math
▸ Bear case — show the assumptions & math
Revenue, capex, free cash flow & debt ($B)
The ultimate fundamental layer. The sheer volume of operating cash flow allows Nvidia to execute $80B buybacks entirely from operational surplus.
The fabless moat visualized: Unlike foundries (TSMC) that require tens of billions in heavy CapEx to manufacture, Nvidia's fabless design model means revenue (sky) and operating cash flow (olive) scale parabolically while capital expenditures (clay) and debt (slate) remain microscopically flat. This operational leverage is why they can return $20B to shareholders in a single quarter without tapping debt. Figures illustrative; fiscal years align primarily with prior calendar years (e.g., FY27 = Cal 2026).
EPS ladder underpinning the targets ($)
The structural integrity behind the price targets. This assumes rapid current compounding slowing toward ~15% terminal growth into the 2030s.
Adjusted (non-GAAP) EPS. The parabolic leap from FY24 to FY26 represents the initial Generative AI and LLM training phase. The base-case (olive) assumes demand continues to compound, but the law of large numbers forces percentage growth down. The $380 base-case target by mid-2031 sits on precisely this ladder: ~$19-21 EPS traded at a normalized ~20× exit multiple.
Scaling at impossible size
Q1 FY27, year-over-year — read these against a $5 Trillion market cap that many label "too big to grow."
Every major metric is compounding at 80%+ off staggering bases. Operating leverage means EPS (+140%) is growing faster than revenue (+85%). Furthermore, frontier segments like Data Center Networking (+199%) demonstrate that Nvidia isn't just selling chips, they are selling the entire interconnected fabric of the data center.
Bull vs. Bear
The entire valuation argument compresses into one structural disagreement: does the hyperscaler spending persist, or are we at peak hardware?
▲ THE BULL CASE
- The Blackwell cycle is just beginning. The GB300 NVL72 ramp is seeing "parabolic" demand from frontier model builders. Supply, not demand, remains the bottleneck.
- It's a platform, not a chip. CUDA software lock-in makes switching to AMD or custom ASICs an agonizing, multi-year engineering hurdle for developers.
- Incredible operating leverage. The 75% gross margin profile allows net income (+214% YoY GAAP) to outrun massive R&D spending, printing $50B+ in quarterly cash flow.
- New buyer pillars. Beyond the 4 major hyperscalers, Nvidia is seeing massive diversification from "Sovereign AI" (nations building proprietary models) and Enterprise/Industrial users.
- Pristine capital return. The board's new $80B share repurchase authorization acts as a massive backstop beneath the stock at any hint of a dip.
▼ THE BEAR CASE
- The Hyperscaler CapEx Cliff. Meta, Microsoft, Google, and Amazon account for nearly 50% of revenue. If their AI applications fail to generate end-user revenue, the music stops instantly.
- Geopolitics and China. Data Center revenue from China has dropped to practically zero. Any conflict involving Taiwan directly interrupts TSMC, halting Nvidia's entire supply chain.
- Custom Silicon (ASICs). Every major customer is frantically designing their own chips (Google TPU, AWS Trainium, Meta MTIA) to escape Nvidia's 75% margin toll booth.
- The law of large numbers. Compounding an $80B/quarter business at 85% is physically impossible to sustain. Multiple compression begins the moment growth slows to "normal" tech levels.
- Margin peak. Gross margins sitting at ~75% can only go down as supply eventually catches up and competitors force pricing pressure on older Hopper architectures.
Risk map — likelihood × impact
Where each risk sits within the node architecture. The hot upper-right corner — likely and high-impact — is the one that matters to the multiple.
- US Export Controls (Edge)
- Growth Deceleration
- Hyperscaler CapEx Cut
- AMD Share Gain
- Custom ASIC Shift
- Taiwan Supply Shock
Hyperscaler CapEx Cut
The top 4 cloud providers decide their AI infrastructure is "built out" for the current generation, pausing orders and crashing NVDA's forward guidance.
Custom ASIC Shift
Google TPUs and Amazon Trainium evolve to handle primary training, breaking Nvidia's monopoly and compressing margins.
Growth Deceleration
Not a crash, but growth falling from 85% to 15%. The business remains immensely profitable, but the stock multiple violently de-rates.
Taiwan Supply Shock
A geopolitical event involving China and Taiwan halts TSMC production, severing Nvidia from its fab supply and crashing global chip output.
AMD Share Gain
AMD's MI-series accelerators capture 10-15% of the data center market by competing aggressively on price.
US Export Controls
Washington bans remaining lower-tier chip sales to specific countries (like the Middle East). Painful, but Nvidia easily sells that supply elsewhere.
The jargon, decoded
Hover the dotted terms in the metrics, or scan the desk's working definitions here.
- Data Center Compute
- The core business segment selling massive GPU clusters (H100, B200) to cloud providers and enterprises for AI training/inference.
- CUDA
- Nvidia's parallel computing platform and API model. It's the "software moat" that makes AI code written for Nvidia practically un-runnable on rival chips.
- Hyperscalers
- The massive cloud infrastructure giants: Microsoft (Azure), Amazon (AWS), Google (GCP), and Meta. They are Nvidia's biggest customers.
- ASIC
- Application-Specific Integrated Circuit. Custom chips built by hyperscalers (like Google's TPU) explicitly to train their own AI, bypassing Nvidia.
- Blackwell (GB300)
- Nvidia's newest generation architecture, succeeding Hopper. Capable of massive performance jumps, its ramp is driving the current guidance.
- Fabless
- Nvidia designs the chips but does not manufacture them. They rely on TSMC (Taiwan Semiconductor) to actually print the silicon.
- Exit multiple
- The P/E assumed at the end of the forecast. Multiply it by projected EPS to get a target price.
- Sovereign AI
- Nation-states buying their own massive Nvidia clusters to train language models on their own localized data and culture.