Profit Over Promise: Why AI Stocks Are Falling and What Comes Next
The AI market is entering a period of financial scrutiny. Investors want evidence of sustainable returns, while founders face a more immediate challenge: building products whose economics work beyond the demonstration.
The slide that said "TBD"
Picture a pitch meeting in 2023.
An AI founder reaches slide fourteen of the presentation. The heading reads, "Path to Profitability."
Below it sits one word: "TBD."
The demonstration has gone well. The model analyses documents, writes reports and generates code. Investors see opportunities across banking, healthcare, insurance and legal services. The company has a promising product and a large addressable market.
The funding arrives.
Nobody spends much time discussing when the company will become profitable.
Now move forward three years.
The product has paying customers. Usage is growing.
Enterprise clients are asking for stronger security controls and deeper integrations. Engineers are spending time improving model accuracy and reducing hallucinations.
Then the monthly infrastructure bill arrives.
Every customer query carries an inference cost. Larger customers require additional support. Smaller customers negotiate on price. Sales cycles take longer than expected.
The founder has built something useful, but the financial picture is becoming harder to explain. Revenue is growing, yet the cost of serving customers is growing alongside it.
This is the problem now confronting the wider AI industry.
For almost two years, investors rewarded companies with exposure to artificial intelligence. Semiconductor manufacturers, cloud providers, data centre operators and software businesses attracted capital on expectations of enormous future demand.
The market is now examining the financial returns behind those expectations.
AI demand is real. The challenge is turning that demand into sustainable profits.
The price of patience has changed
The economics of investment depend partly on the cost of capital.
When interest rates are low, investors have greater room to wait for future returns. A company promising substantial profits several years from now can still attract significant funding.
Higher interest rates change the calculation.
The supplied market research places the 10-year US Treasury yield at approximately 5% in late October 2026, with US inflation reported at 3.7%. At these levels, investors have a stronger reason to compare uncertain future returns with income available from lower-risk assets.
The scale of AI infrastructure spending adds further pressure.
Meta has guided towards capital expenditure of $135 billion to $145 billion for 2026. Major cloud companies are committing hundreds of billions of dollars to computing infrastructure.
These investments support long-term capacity. They also create a financial obligation. Investors need to understand how quickly the spending will translate into revenue and cash flow.
A technology company might establish a leading position in AI while delivering disappointing returns to shareholders. The commercial value of the technology and the financial performance of the company are separate considerations.
That distinction is driving the current correction.
Three markets, three versions of the same problem
The United States, China and India have experienced different forms of AI-related market pressure. Their exposure to the technology also differs.
United States: Infrastructure spending meets financial scrutiny
The American correction initially exposed companies carrying substantial infrastructure and financing costs.
CoreWeave provides a useful example.
According to the figures in the supplied research, its second-quarter 2026 revenue increased 112% to $2.58 billion. The company still reported a quarterly loss of $626 million. Net interest expense reached $640 million.
Its business benefits from demand for GPU computing capacity. The financial structure behind that growth creates a separate challenge.
GPU infrastructure requires substantial upfront investment. Companies must finance hardware purchases and data centre operations before recovering those costs through customer contracts. Debt adds interest expenses, while depreciation affects reported earnings.
Demand alone does not resolve these pressures.
The selling eventually spread to semiconductor and technology stocks. During the week reported on 18 October, the Nasdaq fell approximately 4.5%. The Philadelphia Semiconductor Index recorded its longest losing streak since 2008.
The supplied research also reports declines in Micron, AMD, Intel and Broadcom during the same period. Dell lost approximately 23% over the week.
NVIDIA became a visible example of changing market expectations.
Its market value reached approximately $5.76 trillion on 6 October 2026. Within nine trading sessions, the company lost more than 12% in market value, equivalent to roughly $600 billion.
The supplied research places its trailing price-to-earnings ratio at approximately 27, compared with a five-year median of around 56.
These figures illustrate how quickly valuation expectations can change, even when the underlying business continues to operate.
Another concern involves data centre construction. The supplied research cites reports of at least 47 major US data centre projects being delayed or cancelled during 2026. Analysts reportedly considered only around 35% of the proposed pipeline credible.
A planned facility still needs financing, construction and access to electricity. Until those requirements are met, announced capacity has limited commercial value.
China: Strong technology and difficult valuations
China's AI correction has been particularly severe in some market segments.
The supplied research reports that the STAR 50 index fell approximately 40% from its July 2026 peak, erasing around $480 billion in market value.
The index had previously traded at close to 250 times earnings. Even after the decline, its valuation remained elevated compared with several international technology benchmarks.
The pressure extends beyond semiconductor companies. Chinese technology platforms are investing heavily in AI models and cloud infrastructure while managing weaker economic conditions and regulatory uncertainty.
The commercial challenge is particularly relevant for companies selling access to AI models.
Chinese laboratories have made progress with efficient open-weight models. These models offer businesses more flexibility and often reduce the cost of adoption.
That development benefits customers. It also puts pressure on companies whose business models depend on premium model pricing.
A company needs a clear reason for customers to pay more. Model performance alone becomes harder to monetise when competing systems offer similar capabilities at lower prices.
The decline in Chinese AI stocks therefore reflects financial expectations around the sector. It does not establish that Chinese AI research has failed.
India: Pressure on IT services and growth in infrastructure
India's position is different because its listed market has fewer large companies focused exclusively on AI model development.
Its exposure comes through IT services, engineering, SaaS, telecom and infrastructure.
The pressure on traditional IT services companies is particularly important.
Their established business models often depend on selling engineering expertise and billable hours. AI tools reduce the time required for software development, testing and support.
Clients expect those productivity gains to translate into lower costs.
This creates a difficult commercial adjustment. Service providers need to preserve margins while changing how they price and deliver their work.
The supplied research reports that Infosys reduced its FY27 revenue growth guidance to 1% to 2%. The company reported a 6.2% increase in net profit to ₹7,823 crore, but client expectations around AI-driven cost savings continued to influence the market.
The infrastructure opportunity tells a different story.
Data centres, power suppliers, semiconductor infrastructure and AI server manufacturers stand to benefit from the physical requirements of AI expansion.
The supplied research cites a Goldman Sachs analysis identifying 42 Indian AI-enabler stocks with a combined market value of approximately $670 billion. The group reportedly gained around 60% during 2026 while the Nifty declined.
India therefore faces two simultaneous developments. AI is putting pressure on some established service models while creating demand for new infrastructure capabilities.
The infrastructure opportunity also carries risk. These businesses depend on continued investment and the availability of capital.
The arithmetic behind AI growth
Revenue growth is easy to communicate. Cash generation requires a more detailed examination of the business.
Consider the figures attributed to OpenAI's shareholder documents in the supplied reporting.
The company reportedly generated approximately $5.7 billion in revenue during the first quarter of 2026, around three times the previous year's figure.
Its reported cash burn for the quarter was approximately $3.7 billion.
This means the company spent roughly $1.23 for every dollar of revenue during the period.
There was progress in gross margin, which reportedly improved from 33% to 39%. The company also held more than $73 billion in cash and securities, according to the same reporting.
These figures illustrate the financial requirements of operating a leading AI platform. They do not, by themselves, indicate an immediate financial crisis.
They do show why investors are paying closer attention to the economics of AI services.
Four measurements deserve particular attention.
Inference cost: Every model response requires computing resources. A product serving millions of queries needs to control the cost of generating those responses.
Gross margin: Revenue needs to exceed the direct cost of delivering the service by a sufficient amount. Higher usage is valuable when the business retains an attractive margin.
Customer retention: A customer who continues paying provides stronger evidence of product value than someone who uses a product once.
Free cash flow: After operating expenses and capital expenditure, the business needs to generate enough cash to support its future plans.
These measurements reveal whether growth is strengthening the business.
An AI product might attract customers quickly while becoming more expensive to operate. Founders need to understand the cost of each completed task and the revenue generated by the customer using it.
The same discipline applies to large public companies investing in AI infrastructure.
A lesson from the fibre-optic boom
The current AI correction has a useful historical comparison.
During the late 1990s, investors and telecommunications companies recognised that internet traffic would grow rapidly. Demand for fibre-optic networks was expected to increase across the global economy.
The underlying prediction was correct.
The internet became central to commerce, communication and entertainment. The demand for network capacity continued to expand.
Yet many companies involved in building fibre infrastructure struggled or failed.
Too much capital entered the market. Companies built capacity ahead of demand and assumed that high prices would continue. When supply increased and prices declined, businesses with expensive infrastructure and weak balance sheets faced serious financial pressure.
The technology succeeded. Several of the companies that invested in it did not.
AI infrastructure faces a related challenge.
Cloud providers and semiconductor companies are making large investments because they expect AI demand to grow. Each investment has a commercial rationale. The combined level of spending still needs to produce acceptable returns.
The historical lesson is relevant to investors and founders alike. A growing technology market does not guarantee attractive economics for every company operating within it.
Why profitability is only part of the answer
The market is right to examine profit. A business needs revenue from customers who receive enough value to justify the price.
There is another consideration.
An AI company might report attractive margins while placing significant pressure on the resources supporting its operations.
Data centres consume electricity and water. AI models depend on human-created knowledge. Automation changes employment patterns. Infrastructure requires materials and long-term investment.
These costs do not always appear in a company's financial statements.
A business that ignores them risks creating operational problems for its own future.
This is where the idea of cyclic AI becomes relevant.
The case for cyclic AI
A sustainable AI system should return value to the resources and communities supporting its operation.
The concept of cyclic AI draws on a simple principle. An organisation should account for what it consumes and design ways to renew the systems on which it depends.
Four areas deserve attention.
1. The energy and water loop
AI infrastructure requires electricity and cooling. The International Energy Agency estimated that data centres consumed approximately 415 TWh of electricity in 2024, around 1.5% of global electricity consumption.
Its projections indicate that consumption could approach 950 TWh by 2030, with AI-focused facilities among the fastest-growing sources of demand.
This creates a direct business concern.
Electricity availability influences where data centres are built and how quickly they become operational. Cooling requirements affect operating costs and water consumption.
AI companies need to improve energy efficiency and measure their resource use. Closed-loop cooling and waste-heat recovery deserve greater attention. Access to cleaner power also influences the long-term economics of infrastructure.
For India, the expansion of solar and other renewable energy sources presents an opportunity to support more efficient AI infrastructure.
2. The jobs loop
AI changes the amount of human effort required to complete a task.
For India's IT services industry, this creates immediate commercial pressure. When software development becomes faster, clients expect lower costs or greater output from the same budget.
The longer-term opportunity lies in developing new forms of skilled work around AI.
Companies will need people who understand industry requirements and know how to evaluate AI outputs. They will also need specialists in security, governance and system integration.
Retraining should form part of the productivity strategy. Businesses that reinvest some of their automation gains in people strengthen their ability to adopt and manage new technology.
3. The knowledge loop
AI systems depend on knowledge created by people.
Writers, developers, researchers, artists and domain experts contribute to the information used to train and improve models.
The quality of future AI systems depends partly on the continued production of original human knowledge.
A sustainable system needs to consider consent, attribution and appropriate compensation. These issues influence trust and the long-term availability of high-quality material.
4. The money loop
The financial loop begins with customers.
A business should generate revenue from products that solve meaningful problems. Its pricing needs to cover computing costs, development and ongoing support.
External funding has an important role in research and early development. Over time, customer revenue needs to support a larger share of the company's operations.
This becomes particularly important as model prices decline.
Companies that improve their efficiency as computing costs fall will have greater room to compete on price while protecting their margins.
These four loops connect financial performance with long-term business resilience.
Efficiency in energy use reduces operating costs. Workforce development supports adoption. Responsible knowledge practices strengthen trust. Customer-funded growth gives businesses a more durable financial foundation.
What comes next for AI stocks?
The next phase of the market will depend on financial evidence.
Investors should pay attention to several developments.
Free cash flow: Watch whether AI revenue grows faster than the capital expenditure required to support it.
Financing conditions: Higher interest rates and wider credit spreads increase pressure on companies with debt-funded infrastructure.
Data centre capacity: The difference between announced projects and facilities with secured electricity supply will reveal how much planned investment is becoming operational.
Inference economics: More efficient models and improved computing systems will influence the cost of serving customers.
India's IT services model: The ability of service providers to sell business outcomes and reusable platforms will influence their long-term competitiveness.
China's AI earnings: Model development attracts attention, but commercial revenue and pricing power determine the strength of the businesses behind it.
For founders, these developments offer a practical lesson.
Before expanding an AI product, understand the cost of serving each customer. Measure how much value the product creates. Track retention and gross margin. Build infrastructure plans around actual demand.
A product needs to work reliably in production and justify its place in the customer's budget.
The closing slide, rewritten
Return to the pitch meeting from 2023.
The founder reaches slide fourteen again.
This time, the profitability slide contains actual business evidence.
It shows what customers pay and what the product costs to operate. It explains the computing resources required to complete each workflow. It records the value delivered to customers and the cash generated by the business.
The presentation is less speculative. The company has a clearer path forward.
The AI correction is forcing the industry to examine the financial foundations of its growth. Some companies will grow into their valuations. Others will need to change their pricing, reduce operating costs or focus on more profitable use cases.
The technology will continue developing. The businesses behind it will need to demonstrate commercial discipline.
AI has the potential to improve productivity across almost every industry. Realising that potential requires companies to manage their infrastructure costs and deliver measurable value.
The next generation of successful AI businesses will need to earn customer trust, generate sustainable returns and account for the resources they consume.
The market is asking for profit over promise. The greater opportunity is to build companies whose profits can be sustained year after year.
Sources and notes
The market figures and commentary in this article draw on the research and source references provided in the original drafts.
Forbes, Jim Osman, 18 October 2026.
Business Standard, 28 October 2026.
Free Press Journal and Upstox, 26 October 2026.
Reuters-based reporting on Chinese markets, 30 October 2026.
CNBC, October 2026.
Yahoo Finance, historical market data, October 2026.
Technology.org and Vision Times, 26 to 29 October 2026.
GuruFocus and Intellectia, October 2026.
Goldman Sachs market analysis, as reported by Investing.com.
The Information, reporting on OpenAI financial performance.
Reuters and Sacra, reporting on AI company financial performance.
International Energy Agency, data centre electricity consumption estimates.
Federal Reserve Economic Data (FRED), historical interest-rate and Treasury yield data.
Investopedia, analysis of interest rates and stock market performance.
MarketWatch, economic indicators and market trends.
Several figures originate from analyst estimates, secondary reporting or intraday market data. They are directional and subject to revision. Readers should consult the original publications and current market data before relying on individual figures.
This article is for informational purposes only and is not financial advice. Market data changes frequently, and investors should verify current prices and consult a qualified adviser before making decisions.
Comments