The AI Valuation Paradox: Balancing Hype With Real-World Impact
For the past year, soaring artificial intelligence (AI) startup valuations have been justified by rapid revenue growth, driven by various industries recognizing AI's potential to reshape operations and enhance productivity.
A prime example is OpenAI's valuation, which grew more than tenfold in just three years, from $14 billion in 2021 to $157 billion in 2024, fueled by ChatGPT's success and its impressive projected earnings. The market's confidence in AI is evident in the lofty average revenue multiple of 23.4x commanded by AI startups.
However, we may soon witness a decline in the high funding levels AI startups currently attract, driven by the rise of low-cost, asset-light alternatives. While this in itself is a strong reason for AI startup valuations to deflate, I believe the exaggeration of current AI capabilities leaves room for further corrections.
Recently, we have seen AI startups secure valuations that were thousands of times their annual revenues. For example, xAI and Infinite Reality were valued at $40 billion and $12.25 billion, respectively.
Even considering the growth potential of AI startups, such valuation multiples are excessive. Such outliers can skew data that most AI companies can achieve such heights when the reality is that many more AI startups tend to close their doors before achieving such market success.
AI startups have distinguished themselves from their predecessors with an unprecedented ability to generate revenue. According to Stripe (paywall), today's leading AI startups that have reached an annualized revenue of $30 million have done so five times faster than past SaaS companies.
At the same time, we must acknowledge that AI startups are much more capital-hungry than other tech startups.
OpenAI faces significant operational costs from its flagship product, ChatGPT, spending approximately $700,000 daily (paywall)—over $255 million annually. While these operational costs are offset comfortably by its $3.6 billion annualized revenue (paywall), OpenAI faces intense competition from tech giants such as Google as well as emerging players such as Anthropic.
To maintain its competitive edge, OpenAI must spend an additional $5 billion annually to train new models. This is an expense that will likely continue until OpenAI establishes itself as the undisputed market leader. To put things into perspective, OpenAI's total funds raised stand at $21.9 billion (registration required).
However, recent advancements by new entrants and the limitations of existing AI models cast doubt on both the funding needs and valuations of AI startups.
DeepSeek, the Chinese AI startup, has disrupted the U.S. AI startup ecosystem by demonstrating that premier AI models could be built without exorbitant capital expenditure. Although various experts are disputing this, the company claims that the total training cost was $5.6 million for DeepSeek-R1, the model that delivers performance comparable to OpenAI's ChatGPT.
When we compare the training costs for the two startups, we can see that OpenAI could train new models for less than half a day with DeepSeek's entire budget.
We are already seeing the AI leaders being challenged. After the release of DeepSeek-R1, between January 23 and 25, ChatGPT lost 41.3 million views.
Thus, some investors are questioning if high-performing AI models really cost as much as advertised.
The reasons to believe that AI startups are overvalued are plentiful. Firstly, we haven't yet achieved true artificial general intelligence (AGI), which by definition is capable of performing any intellectual task a human can. What we have right now is a very narrow version of AI that can reliably carry out certain tasks, such as natural language processing or image recognition, but has limited application elsewhere.
Secondly, AI's commercial viability remains questionable. A Boston Consulting Group (BCG) report analyzing 1,000 companies that adopted AI found that only 4% generated substantial value, while only 22% had progressed beyond the proof-of-concept stage to generate any value at all. Notably, the companies that stood out were already well-positioned for success due to strong nonfinancial factors, such as patents filed and employee satisfaction.
Thirdly, various studies note that the capabilities of AI in tasks such as logical reasoning, chemical compound discovery and code writing have been exaggerated.
Thus, only a few AI startups that achieve significant breakthroughs, such as closing the gap between advertised and actual capabilities and enhancing commercial viability, are likely to survive and justify their valuations, while the majority perish.
While some AI startups' values are astronomical multiples of their annual revenues, these cases represent a small group of outliers. Once we exclude the outliers, we can observe reasonable valuation multiples across all stages.
However, a widely recognized cause for concern for AI startups is their struggle to achieve profitability due to their asset-heavy nature and the high costs associated with operations and training.
Additionally, AI's real-world impact remains limited, with narrow applications, questionable commercial viability and sometimes-exaggerated capabilities. As low-cost alternatives emerge, investors are increasingly scrutinizing whether U.S. AI startups can maintain their competitive edge.
The information provided here is not investment, tax or financial advice. You should consult with a licensed professional for advice concerning your specific situation.
Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
The Hill
13 minutes ago
- The Hill
We must loosen China's chokehold on battery supply chains
A ceasefire in the U.S.-China trade war doesn't change the fact that Americans are subject to Beijing's whims when it comes to critical supplies of everything from magnets to minerals. This is not an accident but the result of decades of Beijing's deliberate practices to build monopolies, dominate supply chains, stifle competition, and foster resource dependencies. But the U.S. and its allies can break China's stranglehold on the battery supply chain, if they work together now to build the components and mine the minerals that go into advanced batteries, while fighting back against China's market manipulation. In our new report, Unplugging Beijing: A Playbook to Reclaim America's Advanced Battery Supply Chains, we lay out the scale and scope of China's non-market practices in battery supply chains — dumping, price manipulation, intellectual property theft, monopolies, and forced technology transfers — and, more importantly, say what America can do about it. One key way in which China controls the battery market is through intentional overproduction — making too much of everything — driving prices below profitability in ways that push out competition. For 2025, Chinese analysts are projecting that China will make twice as many electric cars as the entire global demand from last year. While enormous subsidies and state support cushion Chinese companies, American companies cannot sustain unprofitable production. China's decision to dump cheap batteries and underlying minerals on global markets sustains their monopolies but harms free markets and open competition. Beijing may finally be acknowledging that its massive overproduction of just about everything is fueling a race to the bottom. But as the central government frets about what Xi Jinping has labeled 'disorderly price competition,' local governments in China are still backing absurd strategies to juice production, such as state-sponsored programs to sell brand new cars as 'zero-mileage' used cars — sold at a loss and dumped on foreign markets, but allowing companies to inflate sales numbers to justify factories operating at full tilt. While Beijing deploys a suite of non-market tactics at scale, its price manipulation is especially damaging. Advanced batteries depend on a host of refined minerals — lithium, nickel, cobalt, and graphite — that are responsible for most of the cost of the resulting battery. China's intervention in nickel markets, for instance, has saddled Western producers with unsustainable costs. In lithium, Beijing has driven prices up or down at will, undermining competing U.S. projects. To counter this, we propose creating a critical minerals and metals exchange, backed by physical assets and a U.S. strategic stockpile. This would offer offtake guarantees above a price floor to support domestic processors. China's monopolies on mineral processing have also become a weapon in the broader trade war. Beijing has imposed export restrictions on key minerals, including graphite — of which it controls more than 95 percent of global battery-grade processing. To reduce these choke points, we advocate for the creation of special economic zones that co-locate processing, infrastructure, and energy access near known reserves. These zones could take advantage of colocation synergies around large reserves, such as the Salton Sea, and could feature pre-vetted environmental analysis and rigorous safety protocols to localize mining, on-site processing, downstream fabrication, energy, and water needs for all related infrastructure. We also recommend expanding the U.S. Development Finance Corporation's risk appetite to back more processing projects internationally. Beyond supply and demand, China's record on intellectual property theft is extensive. Most Chinese espionage cases involve attempts to acquire commercial technology. The battery sector is a repeated target: the Justice Department has charged Chinese actors with stealing battery tech from Tesla and Phillips 66. Many of China's non-market tactics — from forced labor to environmental shortcuts — thrive in secrecy. To increase transparency, we recommend that the U.S. bar foreign firms from selling into American markets unless they meet strict digital customs and trade data standards. U.S.-listed companies should also be required to map their full supply chains to expose any hidden reliance on forced labor. To compete with all this, the U.S. must invest in cleaner, more efficient, and higher-performing manufacturing processes. We propose increased academic research in battery science in exchange for low-cost licensing to U.S. companies, full cost recovery for research and development in the tax code, and publicly owned modular testing facilities to reduce innovation barriers for smaller firms. There is a way forward — if we choose to act boldly. New supply chains won't emerge from one nation alone. We need domestic reindustrialization and international ally-shoring. Both require upgraded infrastructure and reliable access to the raw inputs of advanced manufacturing — minerals, chemicals, and tooling. Strengthened trade rules, coordinated tariffs, and harmonized regulations among market economies are essential. Most importantly, this effort must be spearheaded by strong American leadership and a dynamic, integrated North American trading bloc. Rebuilding America's supply chains will take industrial work and political will, but we must commit to the hard tasks now to protect our economic security and resilience for the long term. The future of American prosperity depends on it. Elaine Dezenski is senior director and head of the Center on Economic and Financial Power at the Foundation for Defense of Democracies, where Joshua Birenbaum serves as deputy director.
Forbes
14 minutes ago
- Forbes
Monetizing Generative AI In Healthcare: Two Competing Paths
Generative AI is advancing faster than any technology in modern memory. When OpenAI released ChatGPT in late 2022, few in the medical community took notice. Most doctors saw it as a novelty, perhaps useful for administrative tasks or as a basic reference tool, but too unreliable for clinical care. In just a few years, that perception has shifted. Today's GenAI tools from companies like Google, Microsoft and Nvidia outperform most physicians on national medical exams and clinical challenges. Although none are currently recommended for patient use without physician oversight, that restriction is likely to be lifted soon. GenAI capabilities continue to double annually, and OpenAI's GPT-5 model is expected to launch within days. That raises a pressing question: How will generative AI be monetized in healthcare? Two competing visions are emerging. The first follows a familiar playbook: tech companies developing new FDA-approved tools for diagnosis and treatment. The second is clinician-led and would allow patients to use readily available, inexpensive large language models to manage their chronic diseases and assess new symptoms. To understand the advantages of each approach, it's first helpful to examine how different generative AI tools are from the AI applications used in medicine today. The Current Standard: Narrow AI Healthcare's traditional AI tools rely on technologies developed more than 25 years ago. Known as 'narrow AI,' these models are trained on large datasets to solve specific, well-defined problems. These applications are developed by training AI models to compare two datasets, detect dozens of subtle differences between them and assign a probability factor to each. For reliable results, the training data must be objective, accurate and replicable. That's why nearly all narrow AI tools are applied in visual specialties like radiology, pathology and ophthalmology — not cognitive fields that depend on subjective entries in the electronic medical record. Although these tools outperform clinicians, they are 'narrow' and can be used only for the specific task they were designed to complete. Take mammography, for example. Early-stage breast cancer can mimic benign conditions such as fibrocystic disease, causing radiologists to disagree when interpreting the same image. A narrow AI model trained on 10,000 mammograms (half from patients with confirmed cancer and half without) can detect far more differences than the human eye, resulting in 10-20% greater diagnostic accuracy than doctors. To evaluate narrow AI tools, the FDA assesses both the accuracy of the training data and the consistency of the tool's performance, similar to how it approves new drugs. This approach, however, doesn't work for generative AI. The development process and resulting technology are fundamentally different. The New Opportunity: Generative AI Today's large language models (LLMs) like ChatGPT, Claude and Gemini are trained on the near-totality of internet-accessible content, including thousands of medical textbooks and academic journals. This broad, in-depth training enables LLMs to respond to virtually any medical question. But unlike narrow AI, their answers are highly dependent on how users frame their questions, prompt the model and follow up for clarification. That user dependency makes traditional FDA validation, which relies on consistent and repeatable responses, effectively impossible. However, the multimodal nature of generative AI (working across writing, audio, images and video) opens a wide range of opportunities for clinicians. For example, I recently spoke with a surgeon who used ChatGPT to interpret a post-operative chest X-ray. The AI not only provided an accurate assessment, but it did so immediately, a full 12 hours before the radiologist issued a formal report. Had there been a serious problem that the surgeon failed to recognize, the LLM would have proven lifesaving. Even without formal approval, physicians and patients increasingly consult GenAI for medical expertise. While there's still debate over whether LLMs are ready for independent clinical use by patients, the pace of progress is breathtaking. Microsoft recently reported that its Diagnostic Orchestrator tool achieved 85.5% accuracy across 304 New England Journal of Medicine case studies, compared to only 20% for physicians. How Will Generative AI Be Monetized In Medicine? As generative AI becomes more powerful each year, two distinct approaches are emerging for how it might be monetized in healthcare. Each model has trade-offs: Medical care often fails to deliver optimal outcomes at a price patients can afford. This monetization model would seek to close those gaps. Take diabetes, where fewer than half of patients achieve adequate disease control. The result: hundreds of thousands of preventable and costly heart attacks, kidney failure and limb amputations each year. In other cases, patients struggle to access timely, affordable advice for new symptoms or for managing their multiple chronic conditions One way to solve these problems would be through GenAI tools built by entrepreneurial companies. Venture funding would allow startups to apply a process called distillation to extract domain-specific knowledge from open-source foundation models like DeepSeek or Meta's LLaMA. They would then refine these models, training them on radiologist interpretations of thousands of X-rays, transcripts from patient advice centers or anonymized recordings of conversations with patients. The result: specialized generative AI tools designed to address specific gaps in care delivery. Developing these tools will be expensive, both in terms of model training and the FDA approval process. Companies would also face legal liability for adverse outcomes. Still, if successful, these tools would likely command high prices and deliver substantial profits. Although less profitable than an FDA-approved device or bot, the second path to GenAI monetization would prove easier to develop, much less expensive and far more transformative. Rather than developing new software, this approach empowers patients to use existing large language models (ChatGPT, Gemini or Claude) to access similar expertise. In this model, clinicians, educators or national specialty societies — not private companies — would take the lead and share in the benefits. They would create low-cost instructional materials for various patient-learning preferences ranging from digital guides and printed pamphlets to YouTube videos and short training courses. These resources would teach patients how to use any of the publicly available large language models safely and effectively. Using these educational tools, patients would learn how to enter blood glucose readings from home monitors or blood pressure measurements from electronic cuffs. The large language model would then assess their clinical progress and suggest whether medication adjustments might be needed. Patients could also ask questions about new symptoms, receive likely diagnoses and learn when to seek immediate medical care. In conjunction with their physician, they could use GenAI's image and video capabilities to identify signs of wound infections and alert their surgeons. Those individuals with chronic heart failure could monitor their condition more closely and catch signs of decompensation early — allowing cardiologists to intervene before hospitalization becomes necessary. A GI specialist could identify complex intestinal conditions using daily patient-reported inputs, or a neurologist could diagnose ALS by analyzing videos of a patient's gait. Unlike startup models that require tens of millions in funding and FDA approval, these educational tools could be developed and deployed quickly by doctors and other clinicians. Because they teach patients how to use existing tools rather than offer direct medical advice, they would avoid many regulatory burdens and face reduced legal liability. And with 40% of physicians already working part-time or in gig roles, there are hundreds, or likely thousands, of experts ready to contribute. And since generative AI tools can provide information in dozens of languages and literacy levels, they would offer unprecedented accessibility. Which Path Will We Take? The two models aren't mutually exclusive, and both will likely shape the future of medicine. Given the potential for massive financial return from the first approach, we can assume that dozens of entrepreneurs are already developing disease-focused generative AI tools. But rather than waiting for technology companies to introduce GenAI tools, physicians working alone or in conjunction with educational companies have the opportunity to drive the process and improve our nation's health.
Forbes
an hour ago
- Forbes
How To Win At AI-Driven Search: 7 Game-Changing Tips For 2025
Uri Samet is the CEO of Buzz Dealer, a digital marketing and digital PR firm specializing in online reputation management, SEO & ASO. Conversational AI models such as ChatGPT, Grok, Copilot, DeepSeek, Claude, Perplexity and Gemini are transforming how audiences access information. Generative engine optimization (GEO), sometimes referred to as AI optimization (AIO) or generative search optimization (GSO) in industry discussions, is critical for executives and entrepreneurs aiming to stay visible. Unlike traditional SEO, which focuses on ranking higher in search engine results, GEO is about ensuring content is selected as the definitive answer to user queries on AI platforms. GEO has two key approaches: The first involves optimizing content—much like SEO—across online channels, such as websites, social media, blogs and media platforms, for inclusion in AI responses. The second focuses on brand management, similar to online reputation management (ORM), ensuring AI delivers accurate and positive information about a business. To thrive in this AI-driven landscape, businesses need actionable strategies that leverage the unique capabilities of conversational AI models. The following seven strategies can position businesses to lead in AI-driven search and gain a competitive edge in visibility and engagement. 1. Craft Conversational Content To increase your chances of being featured in AI responses, create content that feels natural and engaging, as conversational AI models typically provide clear, user-focused answers. Include relevant statistics, such as data on industry trends or research, to add authority to your answers and thus improve content visibility. Write conversational-style content, but aim for clear, user-focused answers. 2. Target Long-Tail, Question-Based Keywords On AI platforms, users tend to ask specific questions like 'What are the best running shoes for flat feet?' instead of using broad search terms like 'running shoes.' Create content addressing these queries to align with AI conversational models, which excel at answering precise questions. This strategy strengthens GEO by ensuring your content meets AI models' expectations for relevance. Writing blog posts or social media content that directly answers user questions can boost traffic by matching user intent. 3. Incorporate Multimedia For Impact According to Search Engine Land, enriching your content with multimedia such as images, videos and audio can help you stand out in AI-driven search, which AI conversational models prioritize for comprehensive answers. Adding videos with clear, descriptive captions can enhance content appeal, which may make it more likely to be selected by AI models. Explainer videos and infographics can be a great way to maximize your content's impact and reach. 4. Update Content Regularly For Relevance To keep content relevant in AI-driven search, refresh it with current information, as conversational AI models "prioritize content that is well-organized, easy to parse, and dense with meaning." You could consider updating blog posts with new data or trends, for example, to help you stand out as a fresh source. 5. Ensure Consistent Brand Messaging For brand-focused GEO, consistency is critical to shaping how AI models represent your business. Maintaining a unified voice across platforms can increase positive AI mentions, according to the same Search Engine Land piece. On the flip side, inconsistent messaging can confuse AI, resulting in inaccurate responses. First, clearly define your brand's personality and key messages, then document them in a simple guide for all team members for easy reference. Don't forget to regularly review all external content, from social posts to blog entries, against these guidelines to ensure consistent voice and accurate brand representation. Make sure you align your tone, values and facts across websites, social media and directories like Google Business Profile. 6. Participate In Online Communities Engaging on platforms like Reddit and Quora can shape how conversational AI models present your brand, as they pull insights from communities like these. Consider sharing your expertise in discussions to prompt AI models to highlight your brand's strengths in answers. 7. Embrace Real-Time Data And Personalization Conversational AI models rely on fresh data and personalization to deliver relevant answers. Including current information, like customer reviews or event updates, keeps content engaging. This information should be integrated directly into your brand's primary online assets like website information (e.g., updates/news section), blog posts, social media profiles, etc. Any official brand asset that's crawlable and indexable by AI engines would be fit. Creating content tailored to user interests, such as location-specific guides, can boost traffic by aligning with AI's focus on personalized results. Stay Visible Or Get Left Behind GEO isn't just industry jargon. It is already reshaping how brands are discovered, trusted and chosen across AI-driven platforms. Businesses that act now will influence how AI models present their story. Those who hesitate risk being invisible when it matters most. Winning in this new search environment means more than showing up. It means building content that speaks the language of AI, staying consistent across every platform and keeping information fresh. Follow these seven strategies to secure your place in AI-driven search and turn visibility into a competitive advantage. Forbes Agency Council is an invitation-only community for executives in successful public relations, media strategy, creative and advertising agencies. Do I qualify?
