Sepgra asked the most popular Artificial Intelligence engines the following question:
How artificial are the valuations of tech companies with AI ventures?
Here are the responses:
Google Gemini:
The question of whether AI company valuations are “artificial” is currently a subject of intense debate among financial analysts and industry leaders. While some experts point to strong fundamentals and real revenue growth for certain giants, others warn of “smoke and mirrors” and speculative excess that mirror past bubbles.
Evidence of “Artificial” or Inflated Valuations
Several indicators suggest that many AI valuations may be detached from current financial reality:
- The AI Valuation Paradox: Startups with minimal or even zero revenue are receiving valuations in the hundreds of millions or billions based purely on future potential and technical milestones rather than profits.
- Circular Financing: Concerns have arisen regarding “circular flows” where big tech firms invest in AI startups that then use that capital to buy cloud services or chips from the same investors, potentially inflating the perceived market value of both.
- Extreme Multiples: Some companies trade at “eye-watering” price-to-earnings (P/E) ratios; for instance, Palantir’s P/E was recently reported at nearly 400.
- “AI-Powered” Branding: Much like the dot-com era, many companies are adding “AI-powered” to their pitches to multiply valuations overnight, regardless of whether they have a sustainable competitive advantage.
- Low Success Rates: Despite trillions in investment, reports indicate that up to 95% of AI business pilots have failed to provide a return on investment as of late 2025.
Arguments for Rational Valuations
Conversely, some argue these valuations are justified by a fundamental shift in the global economy:
- Real Profitability of Leaders: Major players like Nvidia, Microsoft, and Alphabet have posted steady earnings growth that supports their higher valuations. For example, Nvidia reached a $5 trillion valuation in late 2025, but its P/E ratio remained at a relatively modest 45 times earnings due to its massive profits.
- Self-Financing: Unlike the dot-com bubble, which was fueled by debt, much of today’s AI infrastructure is funded through corporate cash and retained earnings, making the sector more resilient.
The question of whether AI company valuations are “artificial” is currently a subject of intense debate among financial analysts and industry leaders. While some experts point to strong fundamentals and real revenue growth for certain giants, others warn of “smoke and mirrors” and speculative excess that mirror past bubbles.
Evidence of “Artificial” or Inflated Valuations
Several indicators suggest that many AI valuations may be detached from current financial reality:
- The AI Valuation Paradox: Startups with minimal or even zero revenue are receiving valuations in the hundreds of millions or billions based purely on future potential and technical milestones rather than profits.
- Circular Financing: Concerns have arisen regarding “circular flows” where big tech firms invest in AI startups that then use that capital to buy cloud services or chips from the same investors, potentially inflating the perceived market value of both.
- Extreme Multiples: Some companies trade at “eye-watering” price-to-earnings (P/E) ratios; for instance, Palantir’s P/E was recently reported at nearly 400.
- “AI-Powered” Branding: Much like the dot-com era, many companies are adding “AI-powered” to their pitches to multiply valuations overnight, regardless of whether they have a sustainable competitive advantage.
- Low Success Rates: Despite trillions in investment, reports indicate that up to 95% of AI business pilots have failed to provide a return on investment as of late 2025.
Arguments for Rational Valuations
Conversely, some argue these valuations are justified by a fundamental shift in the global economy:
- Real Profitability of Leaders: Major players like Nvidia, Microsoft, and Alphabet have posted steady earnings growth that supports their higher valuations. For example, Nvidia reached a $5 trillion valuation in late 2025, but its P/E ratio remained at a relatively modest 45 times earnings due to its massive profits.
- Self-Financing: Unlike the dot-com bubble, which was fueled by debt, much of today’s AI infrastructure is funded through corporate cash and retained earnings, making the sector more resilient.
- Infrastructure Moats: Companies building the “plumbing” of AI—data centers, chips, and foundational models—are seen as having rational valuations because they provide the essential infrastructure for the entire industry.
Expert Outlook for 2026
As of January 2026, the market is entering a “stress test” phase:
- Correction Fears: Some analysts suggest an AI bubble burst could be imminent, potentially wiping out millions of jobs if the technology falls short of market expectations.
- Selective Survival: Industry figures like Bill Gates have warned that while AI is profound, a “reasonable percentage” of current stocks will likely lose value as competition increases and companies struggle to justify their high “bills”.
Grok
The valuations of tech companies involved in AI ventures are significantly inflated in many cases as of early 2026, blending genuine transformative potential with elements of speculation, hype, and financial engineering that critics describe as “artificial” or bubble-like. While AI is driving real revenue growth (especially for infrastructure players like Nvidia), many valuations far exceed fundamentals like current revenue, profitability, or proven ROI, relying instead on future promises, circular financing, and narrative momentum.
Key Evidence of Artificial/Inflated Valuations (as of late 2025 / early 2026)
- Extreme Market Concentration and Multiples
The “Magnificent 7” (including Nvidia, Microsoft, Meta, etc.) drove ~30% of the S&P 500’s value in late 2025, the highest concentration in decades. US equities traded at ~23x forward earnings — stretched since the dot-com era. AI-specific multiples were even more extreme:
- Median revenue multiples for AI startups/firms: 25–30x EV/Revenue (some categories like LLM vendors hit 35–50x).
- Traditional SaaS: 6–8x (e.g., Spotify ~7x).
- Public AI-linked stocks (e.g., Palantir) traded at premiums implying 45%+ annual growth needed for years to justify prices.
- Major Players’ Valuations vs. Reality
- Nvidia: Market cap reached $4.5–5 trillion (surpassing Germany’s GDP). Fueled by AI chip demand, but tied to circular deals (e.g., $100B investment in OpenAI, expecting reciprocal GPU purchases). Analysts note this creates “circular financing” where money flows back to Nvidia via customer spending.
- OpenAI: Valued at $500 billion (up from $157B in 2024) after secondary sales and deals. Projected 2025 revenue ~$13–20B, but massive losses (~$5B in prior years) and $1.4T future data-center commitments. Critics highlight “vibe revenue” — hype-driven growth without proportional profitability.
- Anthropic: Valued ~$183B after 2025 raises, with similar cash-burn issues (billions in losses).
- xAI (Elon Musk): Rapid raises pushing toward $230B+ valuations despite early-stage status.
- Circular Deals and Financial Engineering
A major “artificial” element: Nvidia invests billions in OpenAI/Anthropic/CoreWeave → those firms buy Nvidia GPUs/data-center capacity → revenue flows back to Nvidia. This closed-loop financing inflates valuations without external demand validation. Similar patterns echo dot-com era cross-investments. Bank of England and others warned of correction risks if infrastructure costs outpace returns. - Industry Voices on Bubble Concerns
- OpenAI’s Sam Altman admitted bubble elements in 2025.
- Anthropic’s Dario Amodei and others flagged “exaggerated” valuations and “cone of uncertainty” around spending.
- CEOs at Web Summit (e.g., DeepL, Picsart) warned of “signs of a bubble on the horizon.”
- Investors (GIC, Goldman Sachs skeptics) called early-stage deals “frothy” despite enterprise demand.
Counterarguments: Not Purely Artificial
- Real Growth Exists: AI infrastructure spending exploded (projected $1.48T–$2T globally in 2025–2026). Nvidia’s revenue surged from chip demand; some firms (e.g., ElevenLabs hitting $200M ARR) show fast scaling.
- Fundamentals Differ from Dot-Com: Many incumbents have strong cash flows and proven ROI (unlike 1999–2000 pure-speculation plays). Goldman Sachs argued gains are backed by earnings, not just hype.
- Adoption Momentum: Enterprise use grows (though slower than consumer), and AI drives productivity in niches (e.g., healthcare, coding).
Bottom Line: How Artificial?
Moderately to highly artificial in the speculative/private startup layer (25–50x multiples on modest revenue, circular deals, loss-making “unicorns” at $100B+).
Less artificial for public infrastructure leaders like Nvidia (real revenue explosion, but vulnerable to slowdowns in AI capex).
The market shows bubble hallmarks: concentration, hype-driven multiples, warnings from insiders, and reliance on future promises over current cash flows. A correction could hit if ROI disappoints, capex proves unsustainable, or competition (e.g., Chinese models like DeepSeek) erodes pricing power — but if AGI or major breakthroughs arrive, today’s “insane” valuations could look conservative in hindsight.
As of January 2026, it’s a high-conviction bet on the future — not a settled fact of value.
The current AI-driven tech valuations (as of early 2026) show notable similarities to the dot-com bubble of the late 1990s–2000, but also significant differences that make a direct repeat less likely — though risks of a sharp correction remain high. The dot-com era saw the Nasdaq Composite surge ~572% from 1995 to its March 2000 peak (5,048), then crash ~78% by late 2002, wiping out trillions and triggering a mild recession. Today’s AI boom has driven massive concentration and multiples, but it’s backed by more tangible revenue and profits in key players.
Key Similarities to the Dot-Com Bubble
| Aspect | Dot-Com Bubble (1995–2000) | Current AI Boom (2023–2026) | Why It Echoes |
|---|---|---|---|
| Hype & Narrative | “Internet will change everything” — endless promises of disruption without proven models. | “AI will transform humanity” — apocalyptic predictions (AGI by 2026–2027) drive FOMO. | Both rely on transformative tech stories amplified by media and easy money. |
| Valuation Extremes | Nasdaq at 70x+ forward earnings peak; many firms at triple-digit multiples with minimal revenue. | Nvidia hit $4–5T market cap; AI startups at 25–50x EV/Revenue; Magnificent 7 at ~30x forward earnings (high but below dot-com peaks). | Extreme concentration: Dot-com had Cisco as world’s most valuable; today Nvidia briefly held that spot. Top 5–7 stocks drive ~30% of S&P 500 (highest in decades). |
| Speculative Capital Flows | VC frenzy, IPO mania; many unprofitable dot-coms went public on hype. | $73B+ VC into AI in early 2025 (58% of global VC); circular deals (Nvidia invests in OpenAI → OpenAI buys Nvidia GPUs). | “Irrational exuberance” — insiders (Altman, Bezos, Dalio) warn of overinvestment and losses. |
| Market Concentration & Risk | Nasdaq crash erased trillions; many firms failed (50%+ public dot-coms gone by 2004). | S&P 500/Nasdaq highs in 2025 driven by AI leaders; IMF, Ray Dalio, and others flag bubble risks comparable to dot-com. | High dependence on few names; reversal could hit GDP (AI capex drove ~half of 2025 US growth per some estimates). |
| Warning Signs | Unprofitable firms dominated; capex outpaced returns. | 95% of GenAI pilots yield zero ROI (MIT 2025); massive debt buildup in hyperscalers. | Circular financing and hype over fundamentals in parts of the ecosystem. |
Key Differences from the Dot-Com Bubble
| Aspect | Dot-Com Bubble | Current AI Boom | Why It’s Different |
|---|---|---|---|
| Revenue & Profitability | Many firms had near-zero revenue (e.g., Pets.com ~$600K before collapse). | Leaders have real, exploding revenue: Nvidia $96B (2025), OpenAI ~$13B ARR, Anthropic targeting $9B+ (2025). | Fundamentals anchor valuations — Goldman Sachs, JPMorgan, Fed Chair Powell argue “this time different” due to profits and business models. |
| Underlying Tech Maturity | Internet infrastructure (fiber) built on speculation; demand lagged. | AI compute demand is real and immediate (data centers, chips); hyperscalers spending $200B+ annually on proven use cases. | AI has enterprise adoption (e.g., coding, healthcare); dot-com was more consumer-speculative. |
| Macro Backdrop | Fed tightening in 1999–2000 triggered burst; high interest rates. | Fed easing in 2025 (rates 4–4.25%); supportive policy. | Less immediate monetary shock risk. |
| Survivors & Legacy | Winners (Amazon, Google) emerged post-crash; internet endured. | Incumbents (Microsoft, Google) reinvesting profits; AI seen as productivity booster. | Potential for lasting value creation — unlike pure speculation. |
| Scale of Investment | VC/IPO frenzy, but smaller absolute dollars. | Trillions in capex projected (e.g., $1–2T global AI infra 2025–2030). | Bigger absolute bets, but on established firms. |
Bottom Line (Early 2026 Perspective)
- Bubble Elements Exist: High concentration, circular deals, hype-driven multiples in private markets/startups, and warnings from insiders (Sam Altman: “yes, there’s a bubble”; Ray Dalio: “very similar”) make parts feel frothy — some analysts call it 17x bigger than dot-com in misallocated capital.
- Not a Pure Repeat: Unlike 2000, leading AI firms have massive real revenue/profits, proven demand, and structural utility. Goldman Sachs/JPMorgan argue it’s “not a bubble… yet” for public leaders; Fed sees “different thing” due to earnings.
- Risk Outlook: A correction (20–50% drawdown in AI stocks) is plausible if ROI disappoints, capex proves unsustainable, or competition (e.g., Chinese models like DeepSeek) erodes pricing. But a full 2000-style crash (78% Nasdaq drop) is less likely — more like an “ebb and flow” with volatility, per Janus Henderson.
- Historical Parallel: Dot-com left the internet; AI will likely leave transformative tools — but the path could involve painful shakeouts for overvalued players.
The market is pricing in enormous future value — whether that’s justified depends on execution over the next 2–5 years. If breakthroughs accelerate (e.g., AGI progress), today’s “insane” valuations could look cheap; if not, expect volatility.
Overview of DeepSeek AI Models
DeepSeek AI is a Chinese artificial intelligence platform specializing in open-source large language models (LLMs) for natural language processing (NLP), computer vision-language tasks, code generation, mathematical reasoning, and more. Founded to advance AI through efficient architectures like Mixture of Experts (MoE) and reinforcement learning, DeepSeek’s models are designed to compete with Western counterparts (e.g., OpenAI’s o1 or GPT series) while emphasizing accessibility, cost-effectiveness, and open-source availability. As of early 2026, their models are hosted on platforms like Hugging Face and accessible via APIs, with a focus on high performance in specialized domains like coding and math. The company provides free access to models like DeepSeek-V3.2 for testing and development.
DeepSeek’s approach includes innovative training methods, such as cold-start data integration before reinforcement learning (RL) and MoE architectures for efficiency. Their models support features like system prompts, JSON output, function calling, and agentic AI use cases, making them versatile for developers and enterprises.
Key DeepSeek Models and Their Details
DeepSeek offers a range of models tailored to different applications. Here’s a breakdown of the most prominent ones based on recent releases (up to late 2025):
| Model | Release Date | Parameters/Context Length | Key Strengths | Use Cases | Benchmarks/Performance Notes |
|---|---|---|---|---|---|
| DeepSeek-V3 | Mid-2025 | 236B parameters; 128K context | General-purpose LLM with MoE for efficiency; supports creative tasks. | Chat, content generation, general NLP. | Versatile for most use cases; outperforms some closed models in open benchmarks. |
| DeepSeek-V3.1 | Late 2025 | Similar to V3; enhanced modes. | Dual-mode (fast/direct like V3 or reasoning like R1); improved versatility. | Switching between quick responses and complex reasoning. | Combines speed and depth; ideal for hybrid applications. |
| DeepSeek-V3.2 | December 2025 | 128K context; max output 4K (up to 8K). | Non-thinking mode; JSON output, function calling. | API integrations, agentic AI, structured data tasks. | Free access available; smooth experience for developers. |
| DeepSeek-R1 | January 2025 | Not specified; focuses on reasoning. | Incorporates cold-start data before RL; strong in math/code. | Reasoning, problem-solving, coding challenges. | Comparable to OpenAI o1 in math, code, and reasoning tasks. |
| DeepSeek-Coder | Earlier variants (pre-2025) | Varies; specialized for code. | Code generation and completion. | Software development, debugging. | Part of the broader suite; excels in programming benchmarks. |
These models are built on principles like efficient training datasets, MoE for reduced computational costs, and RL to enhance reasoning capabilities. DeepSeek-V3 is recommended for general use, while R1 shines in specialized reasoning scenarios.
Training Methods and Mechanisms
DeepSeek’s models use advanced techniques:
- Mixture of Experts (MoE): Distributes computation across specialized “experts” for efficiency, reducing inference costs while maintaining performance.
- Reinforcement Learning (RL): Integrated with cold-start data to improve reasoning without overfitting.
- Data Handling: Focuses on high-quality, diverse datasets for math, code, and multilingual tasks.
- Open-Source Ethos: Models are available on Hugging Face for fine-tuning, promoting community-driven improvements.
This results in models that are cost-effective (e.g., lower GPU requirements) and competitive in benchmarks like MMLU (general knowledge) or HumanEval (coding).
Pricing and Accessibility
DeepSeek offers API access with competitive pricing (as of January 2026):
- Input Tokens: $0.14–$0.70 per million (varies by model).
- Output Tokens: $0.42–$2.10 per million.
- Features: JSON mode, function calling included in newer versions like V3.2.
Free tiers and open-source downloads are available via their website or Hugging Face. For enterprise, they support custom integrations.
Why DeepSeek is Gaining Popularity
DeepSeek models are praised for being open-source alternatives to proprietary giants, with strong performance in non-English languages (e.g., Chinese) and specialized tasks. However, challenges include potential biases in training data and regulatory scrutiny in China. Overall, they’re a go-to for developers seeking affordable, high-capability AI without vendor lock-in.
Open AI (ChatGPT)
Short answer: somewhat artificial—but not totally imaginary.
Longer answer: AI valuations today are a mix of real fundamentals, genuine optionality, and classic hype-cycle behavior.
Here’s how to think about it without drinking the Kool-Aid or becoming a cynic.
1. Why valuations feel “artificial”
📈 Narratives are doing a lot of the lifting
Markets are pricing:
- future dominance, not current revenue
- platform potential, not proven margins
- “AI exposure”, even when it’s thin or vague
If a company can plausibly say “we touch AI,” investors often assume:
“This could be the next AWS / iPhone / Google Search.”
That optionality gets capitalized today, even though realization might be 5–10 years away—or never.
💰 Revenue ≠ profits (yet)
Many AI ventures have:
- explosive top-line growth
- terrible unit economics (compute costs, talent costs, infrastructure)
- unclear long-term pricing power
So multiples are often based on:
- revenue growth rates
- TAM stories
- comparisons to winners (NVIDIA, OpenAI, Google)
…rather than sustainable cash flow.
🧠 AI is a general-purpose technology
This is key. AI is closer to:
- electricity
- the internet
- cloud computing
That makes it everywhere—and also means:
- most value may flow to a few infrastructure winners
- many application-layer companies will be commoditized
Markets are currently pricing many companies as potential winners, even though history says most won’t be.
2. Why it’s not pure bubble logic
🧩 Real productivity gains are already visible
Unlike some past hype cycles:
- AI is demonstrably improving software dev, design, research, ops
- Enterprises are actually deploying it
- Switching costs and data moats are real in some niches
This isn’t pets.com energy.
🏗️ Infrastructure players deserve a premium
Companies controlling:
- GPUs / chips
- cloud compute
- model platforms
- data pipelines
…are closer to “picks and shovels.” Their valuations are often aggressive but defensible, assuming demand continues.
🧮 Optionality has value—just not infinite value
AI creates asymmetric upside:
- one breakout product can justify years of losses
- small probability × massive payoff = real option value
The problem is markets often overestimate the probability.
3. Where valuations are most “artificial”
🚩 Red flags
- “AI-powered” with no proprietary model, data, or distribution
- Heavy reliance on OpenAI / Anthropic APIs with no moat
- Valuations assuming monopoly-like margins in competitive spaces
- Narratives that jump straight from demo → world domination
These look a lot like:
SaaS 2021 + crypto 2017 + dot-com 1999
but with better tech.
4. A useful mental model
Think of today’s AI valuations as:
- 30–40% real fundamentals
- 30% legitimate long-term optionality
- 30–40% speculation, FOMO, and narrative momentum
That mix shifts by company.
Some will grow into their valuations.
Many will stagnate.
A few will implode.
A handful will justify everything.
5. Bottom line
- Are AI tech valuations inflated? Often, yes.
- Are they totally fake? No.
- Is the market bad at distinguishing winners early? Historically awful.
If you want, I can:
- break this down by Big Tech vs startups
- compare it to dot-com or cloud-era bubbles
- or help you evaluate a specific company’s AI valuation
This split is exactly where things get interesting.
I’ll keep it clean, a little opinionated, and grounded in how markets actually behave.
1. Big Tech AI valuations: inflated, but anchored
Who: Microsoft, Google, Amazon, Meta, Apple (to a lesser extent), NVIDIA
Vibe: “Expensive… but you can explain it in a spreadsheet”
Why Big Tech valuations aren’t totally artificial
They already have:
- Massive cash flow to fund AI losses
- Distribution baked in (billions of users, enterprise lock-in)
- Data moats that actually matter for model improvement
- Infrastructure control (cloud, chips, platforms)
So when markets price AI upside, they’re layering it on top of:
very real, very boring, very durable businesses
Example logic investors use:
- “Even if AI only modestly improves margins or retention, that’s worth hundreds of billions in NPV.”
- “Worst case: this company survives. Best case: it owns the next computing paradigm.”
That’s not crazy—just optimistic.
Where Big Tech is overvalued
CapEx risk is underpriced
- AI compute is insanely expensive
- Returns on that spend are still unproven
- Margins could compress before they expand
Internal cannibalization
- AI often replaces existing high-margin products (search, ads, SaaS seats)
- The market tends to price only the upside, not the self-disruption
Winner-takes-most assumptions
- Not every Big Tech firm can be the AI king
- Several are priced like they will be
📌 Net take:
Big Tech AI valuations are stretched but grounded. Think late-cycle cloud boom, not dot-com insanity.
2. AI startups: optionality on steroids
Who: model labs, AI SaaS, vertical AI tools, agents, copilots
Vibe: “This could be worth $0… or $100B”
Why startup valuations feel artificial
Revenue quality is fragile
- Fast growth, low stickiness
- Customers experiment → churn
- Pricing power is mostly theoretical
Moats are thin
- Many are wrappers on foundation models
- Switching costs are low
- Features get copied fast
Valuations assume success and dominance
A $2–10B valuation often implies:
- category leadership
- sustained differentiation
- future margins that look like peak SaaS
That’s a lot of assumptions stacked together.
Where startup valuations do make sense
True differentiation exists when:
- The startup owns proprietary data (medical, legal, industrial)
- The workflow is mission-critical, not “nice to have”
- Distribution is locked in (regulation, compliance, long sales cycles)
- The AI is embedded so deeply that ripping it out breaks the business
These are rare—but real.
3. The biggest valuation mismatch (important)
Big Tech is often under-rewarded for:
- infrastructure risk
- regulatory exposure
- internal cannibalization
Startups are often over-rewarded for:
- demos over durability
- speed over defensibility
- “AI-native” branding over economics
Markets love:
small companies with huge stories
Even when the base rate says most will fail or be absorbed.
4. Quick comparison table
| Dimension | Big Tech | AI Startups |
|---|---|---|
| Cash flow | Massive | Usually negative |
| AI moat | Infra, data, distribution | Often narrow |
| Valuation driver | Incremental upside | Binary outcomes |
| Downside risk | Limited | Extreme |
| Upside realism | Moderate | Often overstated |
| Bubble risk | Medium | High |
5. Bottom line
- Big Tech:
Valuations are ambitious but defensible. You’re paying for durability + optionality. - AI Startups:
Valuations are mostly lottery tickets priced like blue chips.
A few will justify the hype.
Most won’t.
The market is currently pretending it can’t tell the difference.
Alright, big question, classic tech power struggle. Short answer: value migrates over time, and who captures it depends on where differentiation is hardest. Longer, more interesting answer below 👇
1. Chips: brutal moat, lumpy value
Who wins: NVIDIA, TSMC, Apple (silicon), ASML (quietly the kingmaker)
Why they capture value
- Extreme capital intensity + physics limits = few winners
- Performance-per-watt and memory bandwidth are not easily copied
- Once you design around a chip, switching costs are real
Limits
- Cyclical as hell
- Design wins are concentrated → one bad generation hurts
- Long-term, chips tend toward commoditization unless you stay on the frontier
Net: Chips capture outsized value when a new compute paradigm hits (GPUs for AI, ARM for mobile). But it’s feast-or-famine.
2. Cloud: toll booths and gravity
Who wins: AWS, Azure, GCP (and increasingly, infra-adjacent players)
Why they capture value
- Scale economies: cheaper power, better utilization, global reach
- Cloud gravity: once your data + workflows live there, leaving is painful
- They monetize everything: compute, storage, networking, managed services
Limits
- Margin pressure from price wars and custom silicon
- Customers hate lock-in and push back
- Regulators are circling
Net: Cloud captures steady, compounding value. Not the flashiest margins, but insanely durable. Think “infrastructure rent.”
3. Apps: where the margin should be… but often isn’t
Who wins: A small number of category-defining apps (OpenAI, Adobe, Salesforce, Meta, TikTok)
Why they capture value
- User-facing differentiation
- Brand + habit + workflow integration
- Can price on outcomes, not inputs
Limits
- Apps sit on top of clouds → someone else taxes you
- Switching costs are often lower than founders think
- In AI specifically: models/features commoditize fast
Net: Apps can capture the most value if they own the user relationship and a unique data or distribution advantage. Otherwise, margins leak downward.
The deeper pattern (this is the key)
Value tends to flow to the layer that is:
- Most constrained
- Hardest to replicate
- Closest to the bottleneck of progress
Right now in AI:
- Chips capture value because compute is scarce
- Cloud captures value because deployment and scale are painful
- Apps fight for value because differentiation is fragile
Over time:
- Chips → commoditize
- Cloud → becomes regulated utilities
- Apps → if they nail distribution + data, they win big
My hot take 🔥
- Short term (now–2 yrs): Chips win
- Medium term (3–7 yrs): Cloud wins
- Long term: Apps win only if they control the interface and the data loop
Otherwise? The value gets taxed away by the layers below.
Deepseek
Excellent question that cuts to the heart of current market dynamics. The valuations of tech companies with significant AI ventures are widely debated and are considered by many analysts to be highly artificial, inflated, and driven by narrative rather than fundamentals—but with a crucial, self-fulfilling logic.
Here’s a breakdown of why they are seen as “artificial” and the counterarguments.
Factors Contributing to “Artificial” or Inflated Valuations
- Narrative Over Numbers: Valuations are often untethered from traditional metrics like P/E ratios. Companies can see stock prices double or triple with only marginal AI-related revenue. The valuation is based on the option value of a transformative future, not current cash flows.
- FOMO & Speculative Frenzy: Investors (both retail and institutional) are terrified of missing out on “the next platform shift.” This leads to capital flooding into any company with a credible AI story, regardless of near-term profitability. This is a classic speculative bubble characteristic.
- The “Winner-Take-Most” Hypothesis: The market is pricing in a belief that AI will create massive, durable monopolies or oligopolies (like today’s cloud or social media giants). Investors are paying a premium today for the company they believe will be that winner, even if the field is still wide open.
- Extreme Cost & Capex Hype: The enormous capital expenditure (Capex) required for AI (GPUs, data centers, energy) is paradoxically seen as a moat. Companies like NVIDIA, hyperscalers (Azure, AWS, GCP), and even those raising billions to spend on infrastructure are rewarded because the scale of spending is seen as a barrier to entry for others. High spending becomes a bullish signal.
- Earnings Multiplier Effect: A small amount of genuine AI revenue can lead to a massive re-rating of the entire company’s valuation. The market isn’t valuing the current AI revenue stream; it’s valuing the entire future business as being transformed by AI.
The Counterargument: Why They Might Not Be Entirely Artificial
- Genuine Productivity Gains: Early adopters are seeing real, measurable efficiency improvements (in coding, customer service, content creation). If this scales, the potential economic value creation is enormous. Valuations are a bet on this scale.
- First-Mover Advantage & Ecosystem Lock-in: Companies building the leading models (OpenAI, Anthropic), foundational platforms (Microsoft’s Copilot ecosystem, Google’s Gemini integration), or critical infrastructure (NVIDIA’s CUDA) are creating sticky ecosystems. Early leadership can lead to lasting advantages in data accumulation, developer mindshare, and iterative improvement.
- The “Fuel” and “Picks & Shovels” Reality: Some valuations are grounded in very real, near-term financials. NVIDIA is the clearest example—its data center revenue is staggering and growing exponentially. Its valuation, while high, is backed by extraordinary sales and profits. Similarly, cloud providers renting GPU capacity have visible, exploding demand.
- Market Sizing Revisions: Analysts are continuously revising Total Addressable Market (TAM) estimates for AI upward. When you believe a technology will affect every sector and job, pricing a leader at even a small percentage of that future TAM can justify today’s high valuations.
Key Differentiators: Who is More “Artificial”?
- More Artificial (Speculative):
- Pre-revenue AI startups raising at billion-dollar valuations based solely on research team pedigree and a technical paper.
- Legacy tech companies slapping “AI” on their old products and seeing a stock pop without a clear technical edge or monetization strategy.
- “Also-ran” model developers without a clear path to differentiation, sustainability, or profit.
- Less Artificial (Fundamentally Supported):
- Infrastructure Enablers: Companies with proven, soaring sales of AI-critical hardware (NVIDIA) or cloud services (Microsoft Azure, AWS).
- Vertical Integrators with Distribution: Companies like Microsoft that combine a leading model (via OpenAI), a ubiquitous software suite (M365), and a massive cloud platform. The monetization path is clear and already in motion.
- Companies with Proprietary Data & Clear Use-Cases: E.g., Adobe integrating generative AI into its creative suite where it can directly increase subscription value and reduce churn.
Conclusion
Yes, current valuations are largely artificial in a traditional, short-term fundamental sense. They are speculative bets on a hypothetical future economy shaped by AI. This creates a significant risk of a major correction, especially for companies that fail to execute on the narrative.
However, calling them entirely artificial misses the deeper market logic. The market is attempting to price a genuine paradigm shift in its earliest, most uncertain phase. It’s a dynamic where the inflated valuations themselves provide the capital (through stock, debt, and attention) that allows these companies to build the very future they are being valued on.
The ultimate answer will come in the next 2-5 years as the hype cycle settles and we see which companies translate narrative into durable profits and market dominance. The correction will be brutal for the purely artificial, while the few genuine winners may eventually grow into their current valuations.
Microsoft Copilot
Valuations of tech companies with AI ventures are elevated, but not uniformly “artificial.” Some firms are priced on genuine revenue, profitability, and infrastructure demand, while others ride speculative enthusiasm that resembles past bubbles. The reality is a mix of solid fundamentals and pockets of hype.
🧠 Why AI Valuations Look Inflated
Several forces are pushing valuations upward—some grounded, some speculative:
🚀 Drivers rooted in real fundamentals
- Strong profitability among major AI leaders such as Nvidia and cloud hyperscalers. These companies report real revenue growth tied to AI chips, data centers, and enterprise adoption.
- Massive, sustained demand for AI infrastructure, including GPUs, data centers, and model training capacity. This demand is not theoretical—companies are spending billions to meet it.
- Broad industry adoption, from healthcare to finance, which supports long‑term revenue rather than short-lived hype.
💸 Drivers that look more speculative
- Record valuations and deal-making fueled by investor excitement, reminiscent of the dot‑com era. Many analysts warn that enthusiasm may be outpacing fundamentals for some companies.
- High market multiples: While not as extreme as the late 1990s, tech P/E ratios have expanded significantly, and some AI‑linked stocks trade at levels difficult to justify without flawless future execution.
- Soaring valuations for companies with minimal AI revenue, where “AI” is more branding than business model. This is where the most “artificial” inflation tends to occur.
📉 Are we in an AI bubble?
Experts are split:
- Some leaders say no, arguing that today’s companies have real earnings, unlike the dot‑com era.
- Others warn of a bubble, including prominent investors who see parallels to past speculative manias.
- Market data shows mixed signals: AI companies report strong earnings, but overall valuations (e.g., Shiller CAPE) are historically high.
📊 Quick Comparison: Fundamentals vs. Hype
| Factor | Fundamentals‑Driven | Hype‑Driven |
|---|---|---|
| Revenue | Strong, growing (chips, cloud, enterprise AI) | Minimal or none |
| Profitability | High for leaders | Often negative |
| Valuation multiples | Elevated but defensible | Extremely high |
| Investor behavior | Long‑term infrastructure bets | Speculative momentum |
| Risk level | Moderate | High |
🧭 So how “artificial” are valuations?
- For major AI infrastructure players: Not very artificial. Their valuations reflect real demand and earnings.
- For smaller or pivoting companies: More artificial. Some valuations rely heavily on future promises rather than current performance.
- For the market as a whole: Partially inflated. There’s genuine value creation, but also pockets of exuberance that could unwind.
If you want, we can dig into specific companies—Nvidia, OpenAI‑linked firms, cloud providers, or smaller AI startups—to see which ones look overvalued versus fundamentally sound.
Claude (Anthropic)
Tech company AI valuations sit on a spectrum from grounded to speculative, depending on several factors:
More justified valuations:
- Companies with actual AI revenue (Microsoft’s Azure AI, Google Cloud AI services, Nvidia’s chips)
- Businesses integrating AI into existing profitable products (GitHub Copilot, Adobe’s generative tools)
- Firms with defensible moats like proprietary data, distribution, or compute infrastructure
More artificial elements:
- Pure-play AI startups valued on projected adoption curves that may not materialize
- “AI washing” where companies rebrand existing tech as AI to boost valuations
- Multiples based on comparisons to OpenAI/Anthropic that may not apply broadly
- Assumption that current growth rates will continue indefinitely
The frothiness varies by segment. Nvidia’s valuation reflects real demand for chips powering AI workloads. But many smaller companies are priced for scenarios where they become the next platform winner—a low-probability outcome given how many are competing.
What makes this tricky: we’re genuinely in a transformative technology shift, so some high valuations will prove correct in hindsight. The challenge is distinguishing companies building sustainable AI businesses from those riding hype. Traditional metrics like revenue multiples are elevated across the board, but whether that’s a bubble or appropriate pricing for a paradigm shift won’t be clear for years.
The 2021 crypto/Web3 valuations offer a cautionary parallel—real technology, real use cases, but many valuations disconnected from fundamentals.
Meta AI
The valuations of tech companies with ventures in AI can be considered inflated, driven by hype and investor enthusiasm rather than traditional valuation metrics. Here’s what’s contributing to this phenomenon :
- High demand and growth potential: AI startups are valued at a premium due to their potential for rapid growth and disruption in various industries. This is reflected in the median revenue multiple of 29.7x for AI companies, significantly higher than the SaaS industry.
- Investor appetite: The release of generative AI models like ChatGPT has sparked massive investor interest, with venture capitalists and corporations pouring billions into AI startups. This influx of capital has driven up valuations, with some companies reaching unicorn status with pre-money valuations over $1 billion.
- Limited data and benchmarking: Unlike established industries like SaaS, reliable data on AI valuation multiples is scarce, making it challenging to determine accurate valuations. This scarcity contributes to the subjectivity and potential artificial inflation of valuations.
- Speculative investment: Investors are willing to pay a premium for AI startups due to their potential for future growth and market dominance. This speculative nature of investment contributes to inflated valuations.
Some notable examples of AI startups with high valuations include :
- Cursor: Valued at $29.3 billion after raising $2.3 billion in funding
- Luma AI: Valued at $4 billion after raising $900 million in a Series C round
- Unconventional AI: Valued at nearly $4.5 billion after raising a $475 million seed round
- Fal: Valued at over $4.5 billion after raising $140 million in a Series D round
To determine the authenticity of these valuations, it’s essential to look at the underlying fundamentals of each company, such as revenue growth, customer acquisition costs, and retention rates. As the AI industry continues to evolve, it’s likely that valuations will be subject to fluctuations based on market performance and investor sentiment.
Perplexity AI
Many AI‑related tech valuations are stretched and partly driven by hype and easy capital, but they are not uniformly “fake”; some are grounded in real earnings and cash flows, while others are highly speculative and arguably bubble‑like.
What “artificial” means here
When people call these valuations “artificial,” they usually mean one or more of these:
- Prices driven by narrative (AI will change everything) more than by current profits or realistic near‑term cash flows.
- Heavy reliance on venture or corporate capital willing to pay high multiples just to secure exposure to AI.
- Concentration risk: a handful of firms’ AI stories propping up whole indexes.
In that sense, a chunk of the AI sector looks speculative, but not entirely detached from fundamentals.
Public market AI leaders
For mega‑caps (Nvidia, Microsoft, Alphabet, etc.), AI valuations are high but backed by visible revenue:
- In 2025, S&P 500 gains were heavily driven by AI winners like Nvidia, Alphabet, Broadcom and Microsoft, whose AI data‑center and chip revenues grew rapidly.
- Large cloud and platform companies plan around 34% capex increases to roughly 440 billion dollars for AI infrastructure, showing real spending behind the story.
- The index traded at about 23× forward earnings in late 2025, elevated versus history and compared with other markets, which fuels “bubble” concerns but does not mean earnings are imaginary.
So for these firms, valuations are rich but anchored to real demand for AI chips, cloud services, and enterprise software, even if expectations might be too optimistic.
Private AI startups and “artificial” pricing
The most artificial‑looking valuations tend to be in private AI startups and growth rounds:
- By 2024, median seed valuations for AI startups were about 17.9 million dollars, 42% higher than non‑AI startups; Series A and B medians were 30%–50% higher than non‑AI peers.
- Separate analyses show 2024–2025 AI pre‑money medians of about 3.6 million (pre‑seed), 10 million (seed), 45.7 million (Series A), 366.5 million (Series B), and 795.2 million (Series C).
- Median revenue multiples around 25–30× EV/revenue for AI fundraising, with top deals far above that, imply investors pay many years of optimistic growth upfront.
- In 2025, AI firms accounted for roughly 80% of US stock‑market gains, with 30% of the S&P 500’s value held in the five largest companies, reminiscent of past bubbles.
At the extreme, late‑stage AI startups like Anthropic have talked with investors at valuations in the hundreds of billions of dollars with still‑developing business models, which many observers see as classic bubble behavior.
Valuation gap table
How “hyped” vs. “real” they likely are
A useful way to think about it:
- Grounded segment: Profitable or near‑profitable AI infrastructure players (chips, cloud, core enterprise software) with measurable AI revenue and long backlogs look less artificial, even if priced aggressively.
- Speculative but plausible segment: Platform and application companies with strong user traction or data moats but unproven long‑term margins; here, small changes in assumptions can swing fair value a lot.
- Highly artificial segment: Early‑stage or hype‑driven ventures with minimal revenue getting very high valuations mainly because “it’s AI,” large pre‑revenue rounds, and revenue multiples above 30×.
In aggregate, the market around AI clearly shows bubble‑like features (high concentration, extreme multiples, narrative dominance), but whether today’s prices are “too artificial” depends on if AI productivity gains ultimately justify the investment surge, which is still uncertain.
If you tell me whether you’re thinking as an investor, founder, or just philosophically, I can tailor this to how you should act (e.g., portfolio risk, what to ignore, what to watch).
Further Reading: