It is being billed as artificial intelligence’s “Sputnik moment”, the day when the hype bubble burst and American companies in the sector and their investors recognized that other countries wanted to challenge their dominance.

Last month, on January 27, a trillion dollars was wiped off the values of companies in the semiconductor, power and infrastructure sectors exposed to AI, more than half of it from chipmaker Nvidia — the biggest one-day loss in the history of the United States.

Veteran investor Marc Andreessen made the space race-era reference on X following the launch by China’s DeepSeek of R1, an open-source reasoning model that is “on par with” OpenAI’s ChatGPT o1, the successor to the large language model (LLM) that inflated the AI balloon in 2022. The big difference is cost — DeepSeek is potentially 30 to 100 times cheaper to use, according to some claims.

Dubai-based venture capitalist and strategic advisor Kevin Czok, who has built and invested in many travel technology startups, said it is important to recognize that AI isn’t new and has been used in applications like predictive pricing, personalized recommendations and demand forecasting for years. In fact, according to Phocuswright’s Travel Startups Interactive Database, AI has been at the core of almost 400 travel startups since 2005, well before the recent AI hype bubble.

If generative AI costs come crashing down, as seems likely, this number may explode. What sort of landscape will these startups face on launch?

Marcus Nigond, CEO of AI travel startup iWander, a PhocusWire Hot 25 Startup for 2025, said travel startup funding peaked in 2021.

“Right now, we’re still about half 2021 funding but it really picked up again in 2024. The reality is that you have less money, but you can be a lot more efficient with your resources,” he said.

Czok said AI has created “unprecedented funding opportunities” for AI-driven travel startups, but it has exposed a critical divide among investors.

“On one side are those riding the hype wave without a clear understanding of AI’s intricacies or its genuine transformative potential,” he said.

“On the other, more discerning investors derive thoughtful theses by understanding the technology’s nuances and practical applications. Investors with a strong grasp of AI are more likely to identify groundbreaking applications that deliver real-world impact, fueling startups with lasting potential.”

The recent stock market correction will bring barriers to entry down even further.

Andres Martinez, founder and CEO of AI travel planning tool Speakspots, said generative artificial intelligence is one of the two biggest cost elements for startups, the other being geospatial data mapping data.

Martinez said there are hundreds of AI travel planners in the market but most will face problems of scalability.

“We are only seeing a handful of AI travel planners grow and achieve a reasonable scale and when you reach that inflection point the invoices that you get in terms of variable costs are so high compared to your revenue and you are in big trouble and you need to be acquired immediately,” he said.

This is already happening: note Layla’s acquisition of Roam Around in March 2024.

“We have our own data scraping robots and our own database. Our infrastructure costs are only 40% of revenue, so our gross margins are 60%,” Martinez said.

Many of the new travel AI startups will fail or be bought out.

“Some startups might eventually consider an acqui-hire, but those with strong value propositions should aim much higher,” Czok said. “The most successful founders build their business models with sustainability and growth in mind, avoiding reliance on acquisition as their sole exit strategy.”

In a report on the impact of generative AI on travel startups, Michael Coletta, Phocuswright’s senior manager of research and innovation, speculated that “startups building without technical baggage and thinking carefully about how to build for the world as it may look in six or seven years, could gain a distinct advantage over their legacy competition.”

Czok agreed.

“Unlike legacy players, startups can experiment rapidly, test new ideas and fail forward without the risk of disrupting large, established operations. Startups that leverage this freedom to innovate and quickly adapt to changing market dynamics are less likely to burn out, even as incumbents ramp up their AI efforts,” he said.

“The startups that will thrive are those that strike a balance between leveraging partnerships with incumbents to scale quickly, while maintaining a competitive edge through niche focus, agility, and deep innovation. By addressing underserved problems or tailoring solutions for specific demographics, startups can create value in areas the giants often overlook.”

IWander’s Nigond said existing distribution and marketing power are the key advantages that legacy players enjoy.

“You are competing with huge budgets, so you have to be super well-funded. That doesn’t mean that we can’t ride with them one day, either through industry partnerships. I think we’ll see a couple of AI companies turn out to be industry giants in their own right,” he said.

Investors with a strong grasp of AI are more likely to identify groundbreaking applications that deliver real-world impact, fueling startups with lasting potential.

Kevin Czok

Phocuswright’s Travel Innovation and Technology Trends 2025 report also said distribution and technology prowess give the OTAs an advantage.

IWander has built its own AI model for travel that sits on top of LLMs like DeepSeek or ChatGPT. Nigond is not fazed by the competition from big players, either in travel or tech.

“Our metric is how can we create travel content better than ChatGPT or Gemini?” he said. “The difference is that we can take a tone of voice and can make it super-personalized to the end user based off their interests. Google are always going to have a huge amount of data but then we can add some curation as well and make it fun.”

Speakspots’ Martinez said, “Big tech companies like Google, Microsoft and Apple will own the consumer. They will have the AI assistant that the consumer uses on a daily basis. I really don’t understand why there are so many travel companies — large ones and startups — trying to play the game of the AI assistant because they will be dead two to three years from now.”

At the same time, online travel agencies have much of the supply-side of the equation with availability and rates. The sweet spot, said Martinez, is to enable AI-agent to AI-agent communication.

In his 2024 paper, Coletta postulated that as LLMs became capable of handling increasingly sophisticated tasks, startups could be run far leaner, requiring less capital going forward.

Callum McPherson, founder and CEO of Obvlo, a PhocusWire Hot 25 Startup for 2025 that has built a proprietary engine to create AI-generated content for travel companies, agreed.

“To develop models from scratch really does require billions in in investment but to use AI models to build applications, it really doesn’t. It means you need to raise less capital and you can retain more equity and there’s going to be more exit opportunities eventually because you’ve raised less so your valuation doesn’t need to be as high,” he said.

McPherson said a key differentiator from legacy players is having AI-first architecture.

“Our entire architecture is built in a modular way wherever model comes out that’s better than what was there before we can slot it in and start using it.”

The company swapped out its image generation from ChatGPT models to Ideogram and within a day or so the output was better. This could prove to be a crucial differentiator after the emergence of DeepSeek.

“Consumers are becoming increasingly progressive in their expectations, demanding the same level of innovation and personalization from the travel industry [as in other industries]. If legacy players fail to adapt, the field could indeed be leveled for startups that can meet these evolving demands,” Czok said.

It’s worth looking at the numbers for the space race here. NASA estimated in 1985 that the launch of Sputnik cost the Soviets around $33 million in 1985 dollars (around $96 million today). The cost of NASA’s rival Project Vanguard program had reached $110 million in 1957, around $1.25 billion today. Despite that setback, America is seen as the ultimate victor in that race. Let’s not bet on it losing the AI race yet.