In what ways does the rapid rebound of software stocks after the AI disruption sell‑off indicate broader market mispricing of AI integration costs, and how might this influence capital‑allocation strategies across sectors?
The software sector's violent repricing and subsequent partial recovery reveal a market struggling to reconcile two incompatible narratives: the existential threat of AI disruption to incumbent business models, and the persistent difficulty of translating AI investment into measurable returns. This tension has created systematic mispricing across multiple dimensions—overstating near-term disruption risk for some companies while underestimating the true cost and timeline of AI integration for others—with profound implications for capital allocation across the economy.
The iShares Expanded Tech-Software Sector ETF (IGV) experienced a 33% correction from its September 2024 peak near $118 to a February 2025 low of approximately $79, representing one of the sharpest declines in the sector's historyRebound in Software Sector! #IGV #MSFT #PLTR #APP A head and shoulder in #SPX? #XLRE breakout ahead!youtube . This collapse was not a gradual repricing but a cascading crisis of confidence triggered by specific catalysts that forced investors to fundamentally reassess software economics.
The January 2025 release of DeepSeek's R1 reasoning model served as the initial shock. The Chinese startup's claim to have built a competitive large language model for approximately $5.6 million—versus billions spent by American competitors—sent Nvidia shares down 17% in a single session, erasing nearly $600 billion in market value🚨 THIS IS REALLY BAD A catalyst that crashed the stock market in Jan 2025 is coming back. Chinese AI startup DeepSeek is reportedly preparing to release its next-gen model, DeepSeek-V4 really soon. Last year, DeepSeek released DeepSeek-R1 in Jan 2025, and here's how the US stock market reacted: - Nvidia dumped 17% in a day and erased $600 billion. - Nasdaq 100 dropped 3%, wiping out over $1 trillion - Philadelphia Semiconductor index dumped 9.2%, its worst day since the 2020 pandemic. - Other tech companies like Microsoft and Alphabet dropped 5%-6%, wiping out hundreds of billions in the market cap. Talking about the crypto market, over $330 billion was erased as BTC and alts dropped 8%-15% in a day. The reason behind this dump was DeepSeek challenging the belief that cutting-edge AI requires $100B+ in hardware and energy infrastructure. And this time, they are coming up with something even better. As per reports, DeepSeek-V4 is capable of substantially outperforming OpenAI's GPT-5 and Anthropic's Claude 4.5 in repository-level software engineering. It can reportedly invest entire medium-sized GitHub repositories in a single prompt and also enables context windows to exceed 1M tokens compared to just 200K with Claude 4.5 Looking at all this, it seems like once DeepSeek V-4 launches, the next wave of correction will happen in overhyped AI companies, which will bring down both the stock market and the crypto market.x +1. The Nasdaq 100 dropped 3%, wiping out over $1 trillion, while the Philadelphia Semiconductor Index fell 9.2%—its worst day since the 2020 pandemicTHIS COULD SHAKE THE MARKETS AGAIN. The catalyst that rocked global markets in January 2025 is resurfacing — and this time, the stakes may be even higher. Chinese AI startup DeepSeek is reportedly preparing to launch its next-generation model, DeepSeek-V4. The last time the company surprised the market with DeepSeek-R1, the reaction was brutal: • Nvidia plunged 17% in a single session, erasing nearly $600B in market value. • The Nasdaq-100 dropped 3%, wiping out over $1T. • The Philadelphia Semiconductor Index fell 9.2% — its worst day since the 2020 pandemic crash. • Microsoft and Alphabet slid 5–6%, losing hundreds of billions collectively. • Crypto didn’t escape either — over $330B was wiped out as $BTC and major altcoins fell 8–15% in 24 hours. The trigger? DeepSeek challenged the core narrative that frontier AI requires $100B+ in hardware, energy, and massive GPU clusters. It disrupted the assumption that only U.S. tech giants could dominate advanced AI infrastructure. Now, reports suggest DeepSeek-V4 could significantly outperform OpenAI’s GPT-5 and Anthropic’s Claude 4.5 in repository-level software engineering. It is rumored to process entire medium-sized GitHub repositories in a single prompt and support context windows exceeding 1 million tokens — far beyond the ~200K limit seen in competing systems. If these claims prove accurate, the implications are profound. Valuations of hardware-dependent AI giants may face renewed pressure. Semiconductor demand assumptions could be reassessed. And as seen before, when mega-cap tech corrects sharply, liquidity tightens across risk assets — including crypto. Markets are narrative-driven. If DeepSeek-V4 disrupts the AI cost structure again, we may witness another volatility wave — first in AI equities, then cascading into broader indices and digital assets. The key question isn’t whether innovation is coming. It’s whether markets are priced for it.x .
A second wave of selling emerged in early February 2026 following Anthropic's release of Claude Opus 4.6, which demonstrated advanced capabilities in automating legal, financial, and design workflowsIf the recent AI and crypto shocks upset you, you're tracking the wrong cycle | Fortunefortune . The S&P 500 software index dropped nearly 9% in five days, with companies like Thomson Reuters seeing shares plunge over 20%Anthropic is crashing the stock market with their new legal automation plugin Anthropic spooked investors, triggering a sharp selloff as markets feared AI could disrupt software-heavy industries like law and finance. The S&P 500 software index dropped nearly 9% in five days, while companies like Thomson Reuters saw shares plunge over 20% It shows one thing very clearly: Now that AI has reached the real world of work, it will replace countless different SaaS/companies and thus workers.x . Software stocks fell not because they missed earnings, but because long-term cash flow visibility—extending 10 to 30 years—suddenly became uncertainSoftware stocks dropped not because they missed earnings, but due to the future uncertainty coming from AI. Long-term (10-30 years) cash flows are suddenly not that clearly visible. ~ Brad Gerstner (@altcap), founder of Altimeter Capital explains. https://t.co/OT2lUnaFdax .
The speed of the subsequent rebound has been equally dramatic. On February 6, 2026, IGV topped all ETF inflows with $828 million in single-day creations, representing a 12.6% increase in assets under managementDaily ETF Flows: IGV Tops Inflows List - Yahoo Financeyahoo +1. Yet this recovery masks profound disagreement about fair value: IGV remains approximately 35% below its 2024 peak, and 89% of listed software firms now trade below 10x revenue𝐓𝐡𝐞 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐄𝐜𝐨𝐧𝐨𝐦𝐲 𝐢𝐬 𝐚𝐥𝐫𝐞𝐚𝐝𝐲 𝐡𝐞𝐫𝐞, 𝐚𝐧𝐝 @openclaw 𝐢𝐬 𝐪𝐮𝐢𝐞𝐭𝐥𝐲 𝐜𝐡𝐚𝐧𝐠𝐢𝐧𝐠 𝐰𝐡𝐨 𝐦𝐚𝐤𝐞𝐬 𝐦𝐨𝐧𝐞𝐲. Before, #AI was just talking. Now, agents execute. That small shift is forcing a serious repricing across tech, platforms, and even real estate. ➤ 𝐇𝐞𝐫𝐞’𝐬 𝐰𝐡𝐚𝐭 𝐬𝐦𝐚𝐫𝐭 𝐩𝐞𝐨𝐩𝐥𝐞 𝐚𝐫𝐞 𝐧𝐨𝐭𝐢𝐜𝐢𝐧𝐠: Traditional SaaS is under pressure. Agents don’t care about fancy dashboards, they go straight to the backend and finish the job. The market is already reacting: • S&P North American Software Index dropped ~15% in Jan 2026 (worst since 2008) • ~$800B in software market cap wiped in days • 89% of listed software firms now trade below 10× revenue • Average software stock down ~33% This is structural, not temporary. Middleman platforms are also losing their edge. Agents don’t click ads, don’t follow paid rankings, and don’t respect traffic moats. They scan everything and pick the cheapest option instantly. Even Goldman Sachs flagged 2026 as the turning point where aggregators risk degrading into simple data suppliers. Real estate is feeling it too, especially offices. Production is shifting from humans to code. Humans need office space; agents need compute and electricity. The numbers tell the story: • US office prices down ~50% from peak • Vacancy rate already above 20% • Some tech-heavy regions approaching 35% risk • Goldman expects ~20,000 admin/pro jobs lost monthly in the US Capital isn’t disappearing rather it’s rotating. Right now the flow is obvious: compute, energy, and data infrastructure are the new power centers. Energy is becoming the primary AI bottleneck, not chips. Another big unlock is payments. Agents can already compare and decide, but traditional finance still requires SMS codes and human approval. That friction is exactly why programmable crypto settlement is getting attention again. If agents can think but cannot pay, crypto rails become infrastructure. And don’t sleep on “𝐄𝐦𝐛𝐨𝐝𝐢𝐞𝐝 𝐀𝐈”. Once software intelligence is solved, physical execution becomes the bottleneck. Budgets once used for labor outsourcing are shifting into robotics at ~25% annually. Agents are getting bodies, and that expands the opportunity surface massively. For individuals, the play is simple: focus less on manual work and more on orchestration leverage. The edge is moving from typing every line of code to designing systems AI can run. Also watch where AI can decide but cannot fully execute yet, those gaps usually become the next big opportunities. We’re not fully priced into the agent economy yet. But the rotation has already started. 👀x +1.
Three distinct forms of mispricing have emerged from the sell-off and rebound dynamics, each with different implications for investors.
The market's initial reaction treated AI disruption as uniformly threatening to all software companies, but this conflates fundamentally different competitive positions. Companies with proprietary data moats, complex workflow integration, and regulatory barriers face materially different disruption timelines than point solutions. A 2024 Capital One survey found that 73% of enterprise data leaders identified data quality and completeness as the primary barrier to AI success—ranking it above model accuracy, computing costs, and talent shortagesWhy 95% of enterprise AI projects fail to deliver ROI: A data analysis - The Presspresspublications .
Fortune 500 companies leveraging advanced business intelligence platforms registered five times the revenue growth, 89% higher profits, and 2.5x higher valuations compared to industry peersWhy 95% of enterprise AI projects fail to deliver ROI: A data analysis - The Presspresspublications . This suggests the market may be systematically undervaluing software companies with robust data infrastructure while overvaluing those positioned as pure "AI plays" without differentiated data assets.
The most pervasive mispricing involves the true total cost of AI implementation. Organizations routinely fixate on visible direct costs—software licenses and API fees—while ignoring what amounts to an "iceberg" of indirect expenses. Integration services alone can add between £60,000 and £200,000 to initial project budgets, while data governance and compliance activities average between £4,000 and £12,000 annually even for SMEsThe Business Case for AI: How to Calculate ROI Before You Investopenkit .
The ROI timeline mismatch is stark. Deloitte's 2025 survey found that most organizations achieve satisfactory ROI on typical AI use cases within two to four years—significantly longer than the seven-to-twelve-month payback period expected for standard technology investments. Only 6% reported payback in under a yearAI ROI: The paradox of rising investment and elusive returnsdeloitte .
MIT research analyzing 300 AI deployments worth $40 billion found that 95% of enterprise AI projects deliver zero measurable bottom-line impactWhy 95% of enterprise AI projects fail to deliver ROI: A data analysis - The Presspresspublications +1. The Duke CFO Survey tells an even more sobering story: when asked about AI's impact over the past 12 months, the vast majority of CFOs reported "no change" across labor productivity, decision-making speed, customer satisfaction, and time spent on high-value tasksTHE UNCOMFORTABLE TRUTH about AI spending: US tech giants spent $380 billion on AI‑driven infrastructure capex in 2025. But most CFOs still can't point to measurable returns. The latest Duke CFO Survey tells the story nobody wants to hear: When asked about AI's impact over the past 12 months, the vast majority of CFOs reported "no change" across the board. Labor productivity? No change. Decision-making speed? No change. Customer satisfaction? No change. Time spent on high-value tasks? No change. These aren't small companies experimenting with ChatGPT either. 78% of large companies invested in AI during 2025. Real money. Real infrastructure. Real deployments. And most are getting... nothing measurable. The spending is something we've never seen: Microsoft committed $80 billion for fiscal 2025 on AI infrastructure. Google, Amazon, Meta - all racing to match it. Together, the big four hyperscalers will be spending over $380 billion throughout 2025 - 2026. Data centers. GPUs. Power plants. Cooling systems. Real assets. Real cash. Real bets on AI transformation. Meanwhile, in the real economy: Goldman Sachs surveyed their investment bankers about how clients are actually using AI: - 37% of companies are deploying AI - 47% using it to boost productivity/revenue - Only 11% using it to cut headcount So deployment is happening. But Goldman's economists found AI has had "negligible" impact on aggregate labor metrics so far. Their forecast? Measurable GDP impact doesn't start until 2027. Here's where the AI gains ARE showing up: Since ChatGPT launched in November 2022: - Magnificent 7 earnings: Up 145% - Remaining S&P 493 earnings: Up 4% The seven companies building AI infrastructure saw earnings explode. Everyone else got a rounding error. But that performance is about infrastructure BUILD-OUT, not productivity PAYOFF. Nvidia makes chips. Microsoft sells cloud compute. Meta runs models. They're selling picks and shovels in a gold rush. The companies trying to mine gold? Still digging holes. The productivity paradox: US productivity grew 2.3% in 2024. Some quarters hit 2.8%. That's solid - well above the 1.2% average from the 2010s. But here's the problem: that improvement started BEFORE the AI spending boom. Goldman estimates tech contributes 0.3-0.4 percentage points of the productivity acceleration. Meaningful. But not revolutionary. And certainly not enough to justify $380 billion in annual capex. The implementation gap: AI doesn't just plug into existing workflows. It requires: - Complete data infrastructure overhauls - Retraining entire workforces - Redesigning core processes - Managing change across every department Most companies bought the tools, bolted them on, and wondered why nothing changed. Here's the big math problem: $10 billion revenue company spends $100 million on AI. To justify that at 20% hurdle rate: $20 million in annual benefits needed. In reality that means: - Cut 100+ positions, OR - Add $20M revenue at zero marginal cost, OR - Slash 2% from operating expenses Most can't hit those numbers in year one. Many won't hit them in year two. But the Magnificent 7 are priced for transformation happening now. Here's what's coming: Gartner projects total AI infrastructure spending go up to $1.37 trillion in 2026, with hyperscaler capex alone exceeding $600 billion. By 2027, we're looking at $1.75 trillion in AI infrastructure spending globally. The spending isn't slowing down. But if productivity doesn't materialize until 2027-2030 like Goldman projects, we're looking at years of multiple compression in Big Tech. Smart play for investors: Trimming Mag 7 exposure and rotating into small and mid-caps where valuations actually make sense. The S&P 493 trades near decade lows relative to Big Tech, but earnings growth is only 5 points lower. Companies spending trillions today are making 3-5 year bets priced for immediate returns. That's not a trade I'm taking.x .
At the infrastructure level, the capital intensity of AI has created a valuation disconnect that threatens to ripple through the entire software stack. AI data centers show approximately $40 billion in annual depreciation against $15-20 billion in revenue before power and staff costs💸 Little concerning article here at futurism. AI data centers are absorbing huge capital, and a new analysis says the math fails, with $40B yearly depreciation on 2025 builds versus $15-20B revenue. The core pieces age on different clocks, chips churn in 2-4 years, networking around 10 years, buildings far longer, so depreciation snowballs. On those lifetimes, 2025 sites show $40B annual write-downs against $15-20B revenue before power and staff, which already implies negative cash. To earn a normal return at this scale, United States data centers would need about $480B revenue in 2025, far above current run rates. Spending is set at $375B in 2025 and $500B in 2026, so the revenue gap widens as the base grows. For scale, Netflix makes $39B from 300M users, so at similar pricing AI software would need 3.69B paying customers. The point is capital intensity plus short hardware life compress margins unless pricing, load factors, or efficiency rise by about 10x. Chip and construction suppliers win near term, but operators eat depreciation and power risk, and customers see higher prices if providers chase break even. --- futurism .com/data-centers-financial-bubblex . To earn a normal return at current scale, U.S. data centers would need approximately $480 billion in revenue in 2025—far above current run rates💸 Little concerning article here at futurism. AI data centers are absorbing huge capital, and a new analysis says the math fails, with $40B yearly depreciation on 2025 builds versus $15-20B revenue. The core pieces age on different clocks, chips churn in 2-4 years, networking around 10 years, buildings far longer, so depreciation snowballs. On those lifetimes, 2025 sites show $40B annual write-downs against $15-20B revenue before power and staff, which already implies negative cash. To earn a normal return at this scale, United States data centers would need about $480B revenue in 2025, far above current run rates. Spending is set at $375B in 2025 and $500B in 2026, so the revenue gap widens as the base grows. For scale, Netflix makes $39B from 300M users, so at similar pricing AI software would need 3.69B paying customers. The point is capital intensity plus short hardware life compress margins unless pricing, load factors, or efficiency rise by about 10x. Chip and construction suppliers win near term, but operators eat depreciation and power risk, and customers see higher prices if providers chase break even. --- futurism .com/data-centers-financial-bubblex .
Rothschild downgraded Amazon and Microsoft to neutral, arguing that AI infrastructure requires six times the capital to generate the same economic value as the original cloud shift. For every dollar spent on AI infrastructure, the estimated net present value is just 20 cents, compared to $1.40 for mature traditional cloud investments🦔Rothschild downgraded Amazon and Microsoft to neutral, arguing AI infrastructure requires six times the capital to generate the same economic value as the original cloud shift. For every $1 spent on AI infrastructure, the estimated net present value is just 20 cents, while mature traditional cloud investments generate roughly $1.40 for every dollar spent. The Economics Amazon's AWS asset base grew 3.5 times from $64 billion in 2021 to $224 billion, but operating profit only rose 2.4 times, marking the lowest return levels since 2015. Microsoft generated $17 billion in revenue per gigawatt before Azure turned toward AI workloads, now it's around $11 billion. Amazon added 3.8 gigawatts of capacity implying $38 billion in potential annual revenue, but reported just $4.6 billion in annualized revenue growth. My Take This is the clearest breakdown yet of why the AI spending doesn't work. You're putting in six times more money to get back one-fifth the value. Amazon built enough infrastructure to generate $38 billion in revenue but only $4.6 billion actually showed up, meaning just 12% of their capacity is being used profitably. Microsoft's revenue per unit dropped from $17 billion to $11 billion after pivoting to AI. These aren't startups figuring things out, these are the most sophisticated tech firms on the planet and their returns are getting worse. Everyone keeps saying the monetization will come later, but Amazon's numbers show the infrastructure is already there and running. The customers just aren't willing to pay enough to make it worthwhile. Hedgie🤗x .
Software valuation multiples have undergone a structural reset that predates the AI disruption narrative but has been accelerated by it. The median EV/Revenue multiple for software companies fell from a 2021 peak of 6.7x to just above 3x by early 2023, stabilizing around 2.6x through 2023-2024 before ticking up to 3.1x in H2 2025Software Valuation Multiples: 2015-2025 - Aventis Advisorsaventis-advisors .
By early 2026, the average SaaS EV/Sales multiple stood at approximately 3.3x—far below the 2020 peak above 20x and near a five-year lowThe Software Winter Marks a Structural Reset as AI Rewrites the Rules | Investing.cominvesting . This compression suggests market pessimism may have exceeded what fundamentals justify, as core business metrics—recurring revenue, high retention, and strong margins—remain solid across leading companies.
However, the bifurcation within software is becoming more pronounced. Only a handful of companies growing implied annual recurring revenue by more than 30% year-over-year trade above 20x revenue multiples, and those have seen approximately 70% share price gains over the past 12 monthsThe current state of public software valuation multiples is bleak. 89% of the 100+ companies we track trade at less than 10x NTM (next-twelve-months) revenue. Only 3 companies trade at 20x+ NTM revenue. This is a ~$3T public asset class. The vast majority of these companies aren't growing revenue quickly, and haven't since 2021 (year-over-year revenue growth has been flat to down for 8+ quarters). The median implied ARR growth rate is only 15% YoY across all public software. For context, Anthropic reportedly grew from $1B --> almost $10B last year. Only a few are growing implied ARR by >30% YoY, and those are the ones trading above 20x and have seen almost a 70% share price gain in the past 12 months. So while AI might be replacing some budgets, it was not the root cause. AI is also gaining traction in greenfield use cases. The reality is that very few of these companies have built AI products that customers are willing to pay for, let alone pay a premium for. If / when they do, they should see their revenue growth and valuation multiples increase. But if they don't, revenue growth, valuations, and value creation will remain extremely muted, and they will fade away. It is wartime for these incumbent SaaS vendors to innovate w/ AI.x . The reality is that very few software companies have built AI products that customers are willing to pay for, let alone pay a premium forThe current state of public software valuation multiples is bleak. 89% of the 100+ companies we track trade at less than 10x NTM (next-twelve-months) revenue. Only 3 companies trade at 20x+ NTM revenue. This is a ~$3T public asset class. The vast majority of these companies aren't growing revenue quickly, and haven't since 2021 (year-over-year revenue growth has been flat to down for 8+ quarters). The median implied ARR growth rate is only 15% YoY across all public software. For context, Anthropic reportedly grew from $1B --> almost $10B last year. Only a few are growing implied ARR by >30% YoY, and those are the ones trading above 20x and have seen almost a 70% share price gain in the past 12 months. So while AI might be replacing some budgets, it was not the root cause. AI is also gaining traction in greenfield use cases. The reality is that very few of these companies have built AI products that customers are willing to pay for, let alone pay a premium for. If / when they do, they should see their revenue growth and valuation multiples increase. But if they don't, revenue growth, valuations, and value creation will remain extremely muted, and they will fade away. It is wartime for these incumbent SaaS vendors to innovate w/ AI.x .
The software sell-off and rebound have catalyzed a fundamental reassessment of capital allocation strategies across the economy, with three distinct trends emerging.
A decisive shift toward purchasing rather than building AI capabilities is reshaping enterprise technology investment. In 2024, 47% of AI solutions were built internally versus 53% purchased. By 2025, 76% of AI use cases were purchased rather than built internally2025: The State of Generative AI in the Enterprise | Menlo Venturesmenlovc . This reflects a recognition that internally developed tools are difficult to maintain and frequently don't provide competitive advantage in a space as dynamic as AIHow 100 Enterprise CIOs Are Building and Buying Gen AI in 2025 | Andreessen Horowitza16z .
Enterprise AI spending reached $37 billion in 2025, up from $11.5 billion in 2024—a 3.2x year-over-year increase. The application layer captured $19 billion, more than half of all generative AI spending2025: The State of Generative AI in the Enterprise | Menlo Venturesmenlovc . This represents a massive capital reallocation from internal R&D to third-party vendor contracts and SaaS platformsFuturist Prediction Accuracydanluu .
Different industries are approaching AI investment with varying intensity and strategy. Within the top 25% of AI spenders, companies in healthcare, technology, media and telecom, advanced industries, and agriculture lead the pack. Companies in financial services, energy and materials, consumer goods and retail, and travel and logistics are spending lessSuperagency in the workplace: Empowering people to unlock AI's full potentialmckinsey .
The consumer industry—despite having the second-highest potential for value realization from AI—appears least willing to invest, with only 7% of respondents qualifying in the top spending quartileSuperagency in the workplace: Empowering people to unlock AI's full potentialmckinsey . This disconnect between potential value and actual investment suggests significant mispricing of AI integration benefits in consumer-facing sectors.
AI spending has increased dramatically in traditionally non-tech sectors: 60x in construction and 35x in manufacturing according to corporate spending dataNo one is talking about this, but this is the bull case for AI and economic growth. AI spend is up 60x in construction. 35x in manufacturing on @tryramp. Faster planning and smarter automation will transform these strategic industries. This is what drives US economic + labor productivity growth.x . India's healthcare AI market is projected to reach $44.8 million by 2030, while AI-driven contributions to manufacturing could reach $85-100 billion in incremental growth by 2035🚀 India’s AI Stock Picks: The Ultimate Sectoral Watchlist for 2025 & Beyond IT Services & Core Tech •Tata Elxsi •Persistent Systems •LTIMindtree •Happiest Minds Technologies •Infosys •TCS •HCL Tech •Wipro •Tech Mahindra Manufacturing & Automation •Bosch India •L&T Technology Services Financial Services/BFSI •Oracle Financial Services •Intellect Design Arena Digital, Analytics & Niche Plays •Affle India •Datamatics •Kellton Tech •Saksoft •Zensar Technologies •KPIT Technologies AI could reshape India’s economy in a big way, with an estimated $500-600B boost to GDP by 2030 (NITI Aayog): 🧩 By 2035, the impact will be massive: •Financial Services (BFSI): $50-55B incremental growth; up to 25% of sector GDP from AI. •Manufacturing: $85-100B incremental growth; digital techs may drive 40% of total manufacturing costs by 2025 (was 20% in 2021). •IT Services: Over $400B growth; GenAI to increase IT productivity by 40%+ in 5 years. •Retail/E-commerce: CAGR of 33.7% (2025-30); 35-37% productivity boost forecast. •Healthcare: Market to $44.8M by 2030 (from $12.9M in ‘24), propelled by Gov. schemes like Ayushman Bharat Digital Mission. 💡 Key Moves & Investments: •AI could add $450-500B to GDP by 2025 (Nasscom-EY). •GenAI companies in India bagged $524M in VC by Aug 2025 (Venture Intelligence). •Four core sectors (BFSI, Retail/CPG, Healthcare, Industrials/Auto) = 60% of AI-driven contribution. •Data center capex to hit $50-95B as capacity grows from 1GW to 9GW in seven years. AI is not just hype—India’s next growth leap may be AI-fueled. Staying ahead means watching sectoral adoption rates, productivity trends, and policy tailwinds. The runway for disruption is just getting started. #StocksToBuy #StocksToWatchx .
A clear rotation from software to infrastructure is underway in capital allocation. Goldman Sachs Research notes that investors have rotated away from AI infrastructure companies where operating earnings growth is under pressure and capex is being funded via debt, while rewarding companies demonstrating a clear link between capex and revenuesWhy AI Companies May Invest More than $500 Billion in 2026 | Goldman Sachsgoldmansachs .
The current rebound is not a broad-based revival of software value but a targeted rotation into AI infrastructure—specifically semiconductor and software names that benefit from adoption rather than disruptionPortfolio Strategy Amidst a Technical and Sector Rotation Crossroadsainvest . The combined capex of Amazon, Alphabet, Meta, and Microsoft is expected to surge 70% year-over-year to a record $610 billion in 2026AI-driven investment is making a historic contribution to US economic growth: The contribution of computers and peripheral equipment to real Q4 2025 GDP surged to +0.57 percentage points, the highest in history. This marks the 4th consecutive quarter of above +0.30 percentage point addition to the economic growth. This is also TRIPLE the contribution seen in 2024. Spending on information-processing equipment surged +36.1% last quarte. Meanwhile, the combined CapEx of Amazon, $AMZN, Alphabet, $GOOGL, Meta, $META, and Microsoft, $MSFT, is expected to surge +70% YoY, to a record $610 billion in 2026. The AI boom is accelerating.x .
The AI disruption sell-off has fundamentally changed how institutional investors evaluate software investments. Between 55% and 72% of venture capital firms are now actively relying on AI in their due diligence process to analyze pitch decks, assess markets, and flag risksHow Venture Capital Firms Use AI in Due Diligence | Amit Nath, M.S. posted on the topic | LinkedInlinkedin . AI investments comprised over half (61%, or $258.7 billion) of all VC investment in 2025, doubling from 30% in 2022Full Report: Venture capital investments in artificial intelligence through 2025 | OECDoecd .
Key due diligence questions have shifted from "Is this company using AI?" to more sophisticated inquiries: What percentage of revenue is directly attributable to AI? Is the product a thin wrapper around a well-known GenAI offering? How will costs scale as operations scale? Can the AI actually build something, execute complex tasks, or operate in the real world where correlations don't hold?AI and machine learning due diligence (+ checklist download) | Fast Data Sciencefastdatascience +1
The governance challenge has become paramount. OpenClaw, an autonomous AI agent with 300,000-400,000 users, represents both the capability and the risk: it can automate thousands of emails and complex workflows, but has 512 identified vulnerabilities including eight classified as critical, no audit trails meeting regulatory standards, and no segregation of dutiesOpenClaw: The AI Agent Institutional Investors Need to Understand — But Shouldn't Touch | Institutional Investorinstitutionalinvestor .
The mispricing dynamics create distinct strategic opportunities and risks:
For companies with proprietary data and complex workflows: The market has likely over-discounted disruption risk. Capital One's finding that 73% of enterprises cite data quality as the primary barrier to AI success suggests that data moats remain valuable, even as the market prices them at decade-low multiplesWhy 95% of enterprise AI projects fail to deliver ROI: A data analysis - The Presspresspublications .
For infrastructure and platform providers: The $700 billion in planned 2026 hyperscaler capex creates structural demand for chips, power, and connectivity—but the revenue-cost mismatch at the data center level ($40B depreciation vs. $15-20B revenue) suggests margin compression is inevitable unless utilization or pricing improves dramatically💸 Little concerning article here at futurism. AI data centers are absorbing huge capital, and a new analysis says the math fails, with $40B yearly depreciation on 2025 builds versus $15-20B revenue. The core pieces age on different clocks, chips churn in 2-4 years, networking around 10 years, buildings far longer, so depreciation snowballs. On those lifetimes, 2025 sites show $40B annual write-downs against $15-20B revenue before power and staff, which already implies negative cash. To earn a normal return at this scale, United States data centers would need about $480B revenue in 2025, far above current run rates. Spending is set at $375B in 2025 and $500B in 2026, so the revenue gap widens as the base grows. For scale, Netflix makes $39B from 300M users, so at similar pricing AI software would need 3.69B paying customers. The point is capital intensity plus short hardware life compress margins unless pricing, load factors, or efficiency rise by about 10x. Chip and construction suppliers win near term, but operators eat depreciation and power risk, and customers see higher prices if providers chase break even. --- futurism .com/data-centers-financial-bubblex +1.
For non-technology enterprises: The optimal strategy appears to be buying rather than building. Of executives surveyed, 92% expect to boost AI spending over the next three years, with 55% expecting increases of at least 10% from current levelsSuperagency in the workplace: Empowering people to unlock AI's full potentialmckinsey . Companies that have redesigned core workflows for AI report higher EBIT impact—but only 21% have done soThis blew my mind 🤯 McKinsey just reported that 80% of companies now use AI but only 1% are doing it well. That single number explains why every “AI initiative” you’ve heard about sounds impressive but delivers nothing. Most companies didn’t integrate AI. They *outsourced curiosity.* They bought licenses, ran pilots, and called it innovation. The few that *did* transform they look nothing like the rest. They rebuilt workflows around models, retrained staff to think in prompts not PowerPoints, and put AI governance right next to the CFO. McKinsey’s numbers are brutal: - Only 21% redesigned core workflows for gen-AI. - Just 28% have their CEO personally leading governance. - Those who did are already reporting higher EBIT from gen-AI. Everyone else is still “experimenting.” Translation: burning budget on proofs of concept with no system design, no feedback loops, and no metrics. The lesson’s obvious: AI ROI doesn’t come from the model. It comes from the org that’s willing to *break itself* to use it. The future isn’t AI-native companies. It’s *AI-rewired* ones.x .
For institutional allocators: The technology concentration risk is at historic levels, with the Magnificent 7 representing over a third of the S&P 500The AI valuation bubble is now getting silly | Nils Pratley | The Guardiantheguardian . If the AI narrative shifts, diversification benefits will be limited. Global pension assets reached a record $68.3 trillion in 2025, with aggregate allocation to equities falling nine percentage points over 20 years to 48% of total assetsGlobal pension assets rise by nearly 10%, reaching new high - WTWwtwco . CalPERS' adoption of a Total Portfolio Approach in November 2025 signals that traditional asset allocation frameworks may be inadequate for navigating AI-driven uncertaintyEndowments & Foundations as LPs: GP Guide 2026 - Altssaltss .
The rapid rebound of software stocks after the AI disruption sell-off reveals a market that has not resolved its core uncertainty: whether AI represents a productivity tool that enhances incumbent software businesses or a replacement technology that commoditizes them. The 95% failure rate of enterprise AI projects and the two-to-four-year ROI timeline suggest the market initially overestimated near-term disruptionWhy 95% of enterprise AI projects fail to deliver ROI: A data analysis - The Presspresspublications +1. Yet the structural shift toward AI—with 85% of organizations planning budget increases and enterprise AI ROI measurement pivoting from productivity to direct P&L impact—indicates the transformation is real, merely delayedEnterprise AI ROI Shifts as Agentic Priorities Surge - Futurumfuturumgroup +1.
The companies that will emerge as winners are those that can demonstrate measurable revenue contribution from AI features—not just deployment or usage metrics. OpenAI's revenue growth to $13.1 billion in 2025, tripling year-over-year, shows the demand is realOpenAI has drastically revised its financial forecasts, revealing that the cost of training and running its AI models is spiraling well beyond previous projections. While revenue is growing rapidly, expenses are outpacing it significantly. The company now projects a cumulative cash burn of $665 billion by 2030, which is $111 billion higher than previously estimated. OpenAI does not expect to become cash-flow positive until 2030 (projecting +$39 billion that year). By contrast, rival Anthropic is targeting a break-even point as early as 2028. Despite the massive burn rate, revenue is booming. OpenAI hit $13.1 billion in 2025 (more than tripling previous figures) and expects to hit $30 billion in 2026. The company currently has 910 million weekly active users, fueled by growth from its newer 5.1 and 5.2 models. They are aiming for 2.75 billion weekly users by 2030. Inference costs quadrupled in 2025. This caused OpenAI's adjusted gross margin to drop to 33%, well below its 46% target and the typical 70%+ seen in successful software companies. Training costs alone are expected to hit nearly $440 billion by the end of the decade.x . But inference costs that quadrupled in 2025, dropping gross margins to 33% versus a 46% target, reveal that the economics of foundation models may never look like traditional software economicsOpenAI just revised its financial projections and the numbers are staggering. $665 billion in cumulative cash burn by 2030. That's $111 billion MORE than they previously projected. Here's what the updated numbers look like: 📈 Revenue is real and growing fast: — $13.1B in 2025 (3x+ YoY) — $30B projected for 2026 — 910M weekly active users today, targeting 2.75B by 2030 🔥 But costs are growing even faster: — Inference costs quadrupled in 2025 — Adjusted gross margins dropped to 33% — well below their 46% target and miles from the 70%+ you need in software — Training costs alone are expected to hit $440B by end of decade — Peak annual loss of ~$170B projected in 2028 💰 The path to profitability: — OpenAI doesn't expect to be cash-flow positive until 2030 — By contrast, Anthropic is targeting break-even by 2028 This is the most expensive bet in tech history. Revenue is scaling at an extraordinary rate. But so is the cost to generate that revenue. 33% gross margins in a software business is a structural problem. Traditional SaaS companies operate at 70-80%+. If inference costs don't come down dramatically, the unit economics don't work at any scale. The big question isn't whether AI will be massive. It clearly will be. The question is whether the economics of foundation models can ever look like software economics — or whether this is a fundamentally different business with fundamentally different margins. OpenAI is betting $665 billion that the answer is yes.x .
The market's current state—with 89% of software firms below 10x revenue but institutional capital flooding back into IGV—reflects a bet that the sell-off overshot. Whether that bet pays off depends entirely on whether AI integration costs decline faster than AI disruption accelerates. For capital allocators across sectors, the implication is clear: allocate to companies with demonstrable AI revenue contribution and data moats, underweight pure-play software lacking differentiation, and maintain discipline around the true total cost of AI integration—because the market has not yet priced it correctly.