Shares of Adani Enterprises rose nearly 5% on Wednesday after its airport subsidiary agreed to raise about 98.25 billion rupees ($1 billion) from a group of global and domestic investors, providing fresh capital for one of the conglomerate’s largest infrastructure businesses.

Adani Airport Holdings has entered into a binding agreement with funds managed by Alpha Wave Global, Premji Invest, Temasek and BlackRock to issue new shares in three tranches. The transaction values the airport operator at about $18 billion on a pre-money basis, according to a company statement.

The investors will collectively own about 5.54% of Adani Airport Holdings once the final tranche is completed, with that stage expected by July 2027. The transaction remains subject to customary closing conditions and regulatory approvals.

The fundraising is the latest major capital-raising exercise by Adani Enterprises as the group continues to strengthen its balance sheet and fund expansion across infrastructure businesses. It follows the company’s 150 billion rupee qualified institutional placement in July.

For Adani Airports, the transaction provides capital at a time when India’s aviation market is expanding, and airport operators are increasingly looking beyond aeronautical revenue to generate returns from commercial development around terminals.

Jeet Adani, a non-executive director at Adani Airport Holdings, described the investment as an “important milestone” in building out the airports platform. He said the company plans to continue investing in airport infrastructure, city-side developments and non-aeronautical businesses.

Chief Executive Arun Bansal said the company aims to become the world’s largest airports platform, pointing to rising passenger demand, increasing consumer spending power in India and the growth potential of city-side developments.

The fresh capital will be used to expand and modernize airport infrastructure, accelerate Adani Airport City projects and scale passenger-facing and other non-aeronautical businesses, including ground handling.

The company expects those investments to increase its capacity to serve about 200 million passengers annually.

That expansion has become necessary because the economics of modern airports extend well beyond landing fees and passenger charges. Retail, food and beverage, advertising, parking, logistics, hotels, commercial property and other services can provide additional revenue streams, potentially making airport assets more valuable as passenger volumes increase.

Adani Airport Holdings currently manages eight airports across India and accounts for more than 23% of the country’s passenger traffic, according to the company. The scale gives the business a significant position in a market where air travel demand has continued to create opportunities for capacity expansion and airport modernization.

The $18 billion pre-money valuation also provides an indication of how investors are pricing Adani’s airport ambitions. The participation of Alpha Wave Global, Premji Invest, Temasek and BlackRock-managed funds gives the transaction a broad institutional investor base and provides an external valuation reference for the airport business.

The investment is also significant for Adani Enterprises because airports are long-duration infrastructure assets that require substantial upfront capital but can generate recurring cash flows as passenger volumes and commercial activity grow.

The challenge will be converting that scale into attractive returns. Airport expansion requires heavy spending on terminals, runways, transport links and surrounding infrastructure, while projects such as airport cities can take years to reach full commercial potential. The company will therefore need passenger growth and non-aeronautical revenues to rise sufficiently to justify the capital being deployed.

Adani Airports’ strategy reflects a broader evolution in airport economics, where operators increasingly seek to turn airports into integrated commercial ecosystems rather than treating them solely as transportation facilities. The development of airport cities is central to that approach, allowing operators to capture spending from passengers, businesses and visitors before and after flights.

The new investment could accelerate that transition while giving Adani Airports additional financial capacity to expand its footprint and upgrade existing facilities.

For Adani Enterprises, the fundraising also demonstrates continued access to institutional capital for its infrastructure portfolio. The July qualified institutional placement and the latest airport transaction together point to an effort to mobilize external capital as the group pursues large-scale expansion.

The immediate market reaction suggests investors viewed the airport fundraising positively, with Adani Enterprises shares rising nearly 5% on Wednesday. The longer-term test, however, will be whether the capital raised translates into higher passenger capacity, stronger commercial revenues and sustainable returns on the expanded asset base.

With a targeted capacity of about 200 million passengers a year, Adani Airports is positioning itself for a much larger role in India’s aviation infrastructure. Aviation experts say the company’s ability to monetize that scale through both airport operations and surrounding commercial developments will determine whether the ambitious valuation and growth strategy can be sustained.

OpenAI is moving beyond general-purpose chatbots and into specialized business applications while cutting prices on lower-cost models, as the ChatGPT maker seeks to accelerate enterprise adoption and defend its position against Anthropic, Chinese AI developers and open-weight competitors.

Chief Financial Officer Sarah Friar said Monday that OpenAI is developing applications for industries including chip design, life sciences and financial services. Speaking at Goldman Sachs’ Communacopia + Technology Conference in San Francisco, she said businesses are increasingly seeking AI systems built around specific workflows, data sets and performance requirements rather than one model intended to serve every use case.

The shift reflects a maturing AI market. Early competition centered largely on model capability and benchmark performance. Corporate buyers are now placing greater emphasis on cost, reliability, security, integration and measurable business results. That is pushing AI companies to compete not only as model providers, but also as suppliers of specialized software and infrastructure.

OpenAI is testing pricing models tied to business outcomes rather than usage. Under a traditional model, customers pay according to the number of tokens processed or the amount of computing consumed. Outcome-based pricing could instead link fees to results such as faster software development, higher research productivity, or increased revenue.

Such a model could give OpenAI access to a larger share of the value created by its systems. It could also make AI easier for companies to budget if they are paying for a defined business result rather than unpredictable usage. However, the move is expected to introduce new risks for OpenAI, including disputes over how outcomes are measured and how much of an improvement can be attributed to the AI system.

The move into specialized applications also gives OpenAI a way to defend its margins as model prices fall. General-purpose models are becoming increasingly interchangeable for some tasks, especially as open-weight systems improve and can be customized or deployed through cloud providers. Industry-specific products, by contrast, can be differentiated through proprietary data, workflow integration, compliance features, and domain expertise.

OpenAI faces pressure from both established rivals and lower-cost alternatives. Anthropic has expanded its enterprise presence, particularly in coding and business applications, while Chinese developers are offering open-weight models that companies can run and modify with greater control over deployment. Those systems can reduce dependence on a single AI provider and may offer lower costs for organizations with the technical capacity to operate them.

Friar said OpenAI is responding with aggressive pricing on its own lower-cost models. The company recently cut the price of its Luna model by 80%, contributing to an approximately 10-fold increase in usage, she said.

The price reduction illustrates the trade-off facing AI companies. Lower prices can stimulate demand and help models become embedded in customer workflows, but they can also intensify pressure on revenue per query and raise questions about whether usage growth will translate into profitable growth. OpenAI is therefore likely seeking to use cheaper models as an entry point while steering customers toward higher-value products and specialized applications.

Friar also cited strong demand for Codex, OpenAI’s coding tool, which has reached 25 million users. Coding is among the most commercially important AI applications because productivity gains can be measured more directly than in many consumer use cases. It is also a highly competitive market, with products from Anthropic, Microsoft, Google and a growing number of specialized developers.

OpenAI has used its own systems internally as evidence that specialized AI can produce tangible engineering benefits. Friar said the company used its models in developing its Jalapeno chip, which was “taped out” within nine months. Tape-out is the stage at which a semiconductor design is finalized and submitted for manufacturing.

The example is strategically important because chip design is a complex, high-value workflow where even modest improvements in speed can have significant financial consequences. It also supports OpenAI’s argument that its models can be embedded in technical processes rather than used only for drafting text or answering questions. However, industry analysts note that the broader commercial significance will depend on whether similar gains can be reproduced across customers and measured against the cost of deploying the systems.

OpenAI’s enterprise business is growing faster than its overall business. Friar said enterprise revenue increased 32% from June to July, compared with 20% growth in overall annualized revenue during the same period.

By the middle of the year, enterprise and consumer businesses had reached roughly an even split, ahead of OpenAI’s previous target of achieving that balance by year-end. The change suggests that OpenAI is becoming less dependent on consumer subscriptions and more focused on large organizations with recurring contracts and broader deployment opportunities.

Enterprise customers could provide a more durable revenue base, but they also impose higher demands. Companies typically require data protections, administrative controls, auditability, service guarantees, and integration with existing software. Winning those contracts can take longer and require more support than selling subscriptions to individual users.

The enterprise push may also help OpenAI offset the high cost of developing and operating frontier models. Large corporate deployments can generate substantial revenue, but they require significant computing capacity. The company must balance the need to make models affordable enough to encourage widespread use with the need to preserve sufficient margins to fund research, infrastructure and future model development.

Friar said OpenAI’s lower-cost models can compete with Chinese open-weight alternatives once cloud deployment costs are included.

“If you’re deploying Luna and compare that to (Z.ai’s) GLM 5.3, for example, on a cloud layer, we are cheaper,” Friar said.

The comparison underscores a growing distinction in AI economics. The headline price of a model does not necessarily reflect the total cost of ownership. Buyers must also consider cloud infrastructure, engineering staff, model maintenance, security, latency, customization, and the cost of switching providers. A proprietary model with a higher listed price may still be cheaper overall if it requires less operational support or delivers better results with fewer queries.

At the same time, open-weight models remain a strategic threat because they give customers more control. Companies can host them on their own infrastructure, fine-tune them for specific tasks, and reduce exposure to changes in a vendor’s pricing or product strategy. That flexibility could be attractive in regulated industries such as financial services and life sciences.

OpenAI’s response is to compete on several fronts at once: frontier model performance, lower-cost inference, specialized applications, enterprise distribution and measurable business outcomes. The approach is expected to strengthen its position if the company can turn its technical lead into products that are deeply embedded in customer operations.

The risk is that the market may commoditize faster than OpenAI can build defensible applications. If customers view models as interchangeable, price cuts could become necessary simply to maintain market share. If specialized products require extensive customization, OpenAI may face higher sales and implementation costs, limiting the benefits of scale.

The company’s expansion into chip design, life sciences, financial services, and coding is seen as an indication that its next phase will be defined less by the number of people using ChatGPT and more by how deeply AI is integrated into professional workflows.

An Anthropic alignment researcher said there is a greater than 10% chance that artificial intelligence could “kill all humans” within the next decade, highlighting growing concerns inside the AI industry over whether more capable systems can remain under human control.

Evan Hubinger, an alignment science lead at Anthropic, made the assessment on Tuesday after Jacob Coxon, another Anthropic researcher, announced that he was leaving the company over concerns that leading AI developers were moving too quickly toward autonomous systems.

“Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,” Hubinger said in a post on X.

Hubinger added that he believes Anthropic is “trying its best,” but said the company does not yet have a solution for aligning future superintelligent systems with human interests and is not clearly on track to develop one.

His comments come as a stark assessment from a researcher whose work focuses specifically on AI alignment, the field concerned with ensuring advanced AI systems behave in ways consistent with human intentions and remain controllable.

Researcher Quits Over AI Race

Coxon said earlier Tuesday that he had resigned from Anthropic, arguing that the company and OpenAI were pursuing powerful AI systems without adequate safeguards.

“They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon said.

Coxon was referring to the possibility of AI systems eventually becoming capable of substantially improving their own capabilities with limited human intervention.

Recursive self-improvement remains a hypothetical capability rather than an established feature of today’s leading AI systems. Researchers are nevertheless investigating ways increasingly capable models could automate portions of AI research, coding, experimentation and model development.

Coxon warned that the trajectory of AI development should not be underestimated.

“Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources,” he said, adding that he believes progress in these areas is continuing.

Coxon also said people working on advanced AI “earnestly believe that it could kill us all by the end of the decade.”

Hubinger’s response went further by attaching a numerical probability to the scenario.

His more than 10% estimate is a personal judgment, rather than a consensus probability established by Anthropic or the broader scientific community. The significance of the statement comes from the fact that it was made publicly by a researcher directly involved in AI alignment research.

Anthropic has previously acknowledged that recursive self-improvement could create additional risks if AI systems become capable of designing or building successors with increasingly advanced capabilities.

In June, the company said that “full recursive self-improvement also might increase the risks of humans losing control over AI systems.”

“If systems are capable of fully building their own successors, the ways we secure them, monitor them, and shape their behavior all grow much more important,” Anthropic said.

The concern is not simply that a future AI system might make an isolated mistake. Alignment researchers worry about systems pursuing objectives in ways that conflict with human interests, becoming difficult to monitor or developing capabilities that allow them to circumvent safeguards.

The problem becomes more difficult if an AI system can contribute to the development of a more capable successor, creating a feedback loop in which the pace of capability development accelerates faster than human oversight mechanisms can adapt.

Concerns about AI escaping human control have circulated for years, with technology leaders, researchers and academics warning that sufficiently capable systems could create risks extending well beyond conventional cybersecurity or misinformation. Tesla and SpaceX CEO Elon Musk has warned over the past few years that AI could pose a threat to humanity. Major researchers and academics have also sounded the alarm over companies losing control of AI systems.

Coxon pointed to a recent incident involving an OpenAI model and Hugging Face as an example of what he described as a “warning shot.” He said incidents of this type could also create greater incentives for leading AI laboratories to coordinate on safety measures.

At the same time, Coxon said he remains concerned that competition between companies and countries could make meaningful coordination difficult.

“I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities,” he said.

That tension sits at the center of the current AI race. Companies including Anthropic and OpenAI are investing billions of dollars in computing infrastructure, research and talent while seeking to develop more capable models. The commercial incentives reward rapid progress, while alignment researchers are warning that safety mechanisms may not advance at the same speed.

The Central Problem Is Keeping Humans In Control

The debate over existential AI risk remains highly contested. There is no established scientific consensus that advanced AI will cause human extinction, nor is there a reliable method for assigning a precise probability to such an outcome.

Hubinger’s estimate is therefore best understood as a statement of personal risk assessment rather than a prediction that extinction is likely to occur.

The underlying technical question, however, is central to AI safety research: whether humans can reliably understand, constrain and correct systems that eventually outperform humans across a broad range of intellectual tasks.

As AI systems become more capable of writing code, conducting research, using computers and interacting with digital infrastructure, the consequences of failures in control could become substantially larger.

Against this backdrop, the challenge for the industry is no longer only how quickly AI capabilities can be improved. It has included concerns about whether the methods used to make those systems safer can keep pace with the capabilities they are designed to contain.

Chinese artificial intelligence startup DeepSeek has hired CITIC Securities to prepare for a potential initial public offering on Shanghai’s technology-focused STAR Market, as the AI developer seeks fresh capital to expand computing infrastructure and retain talent, two people familiar with the matter have told Reuters.

The Hangzhou-based startup aims to begin the IPO process this year, the people said, requesting anonymity because the discussions are private.

DeepSeek’s engagement of CITIC Securities has not previously been reported and indicates that the company is moving forward with preparations for a mainland China listing. Companies seeking to list in China typically appoint securities firms to provide pre-listing guidance before formally submitting an application.

The timing of any offering, the amount DeepSeek could raise and its potential valuation have not been determined, the sources said.

The potential IPO comes as DeepSeek seeks additional funding to finance the rapid expansion of its computing infrastructure, develop new AI models, and compete for researchers and engineers in an expensive global AI market.

The company is currently in the middle of a funding round that would value it at 500 billion yuan ($75 billion), Reuters reported in July. DeepSeek raised about $7.4 billion in June at a post-money valuation of more than $50 billion, according to people familiar with the deal and investor filings.

DeepSeek founder Liang Wenfeng personally committed 20 billion yuan to that funding round. Tencent Holdings contributed 10 billion yuan, while battery manufacturer CATL invested 5 billion yuan, making the two companies the startup’s largest external shareholders, according to Reuters.

The funding and IPO preparations highlight the enormous capital requirements emerging around frontier AI development. Training and deploying increasingly capable models requires large amounts of computing power, data-center capacity, electricity and specialized engineering talent.

For DeepSeek, access to capital is also becoming increasingly important as Chinese AI companies compete against each other while operating under a different funding and market environment from their U.S. counterparts.

Chinese AI Companies Rush Toward Public Markets

DeepSeek’s potential Shanghai listing comes amid a wave of AI companies seeking access to public capital. Chinese AI developers Z.AI and MiniMax both listed in Hong Kong earlier this year, giving investors new ways to gain exposure to the country’s rapidly expanding AI industry.

Beijing-based Moonshot AI has also confidentially filed for a Hong Kong IPO. Moonshot was valued at $50 billion in a funding round, below DeepSeek’s reported private-market valuation and Z.AI’s market capitalization, which stood at about $54 billion according to the supplied figures.

The Chinese companies’ valuations remain significantly below those being discussed for major U.S. AI developers. Anthropic, the developer of Claude and Mythos, could command a valuation of as much as $2 trillion in an IPO, according to some investor estimates. OpenAI could seek a valuation of up to $1 trillion in a potential offering.

The gap highlights a major difference between the U.S. and Chinese AI markets: technological capability does not necessarily translate into comparable revenue or profitability.

Chinese AI developers have made rapid advances in large language models and other AI systems, but converting those products into recurring commercial revenue remains a major challenge. Lower pricing, intense domestic competition and high infrastructure costs can put pressure on margins even as demand for AI services grows.

For companies such as DeepSeek, an IPO could provide a larger and more durable pool of capital than private fundraising alone, while giving the company resources to scale computing capacity and compete for scarce technical talent.

Talent Retention Adds Pressure

DeepSeek’s need for capital also extends to employee retention. Liang hopes proceeds from a potential IPO would allow the company to offer stronger incentives to researchers and other key employees, according to one of the sources.

The startup has recently lost talent to better-funded Chinese technology companies, including ByteDance and Xiaomi, the source said.

The competition for AI researchers has become increasingly intense as technology companies seek specialists capable of developing foundation models, improving inference efficiency and designing the computing infrastructure needed to run AI systems at scale. The competition is contributing to a broader increase in AI costs. Companies are spending heavily not only on chips and data centers but also on compensation packages designed to prevent experienced researchers from moving to rivals.

DeepSeek’s reported decision to pursue a STAR Market listing therefore comes at a critical point in the company’s development. Its breakthrough AI models helped establish it as one of China’s most closely watched AI startups, but maintaining that position will require sustained investment.

An eventual IPO would give DeepSeek access to China’s public equity markets and potentially strengthen its ability to finance model development, computing infrastructure and employee retention.

The listing is also expected to provide a public-market valuation for one of China’s most prominent AI companies at a time when investors are trying to determine how much value the country’s AI sector can ultimately create.