What AI Video Means for the Future of Sports
AI-generated video is advancing rapidly. That has major consequences for the sports industry.
Last week, I saw an AI-generated video showing Lionel Messi, Cristiano Ronaldo, and other soccer stars meeting their younger selves. Check it out. It was a fun 23 seconds.
But beyond fun, the video felt significant. This was the first AI-generated sports content to bring me a serious amount of entertainment value.
It got me thinking: What does the rise of AI-generated video mean for the future of sports?
Over the past two years, both startups and Big Tech have made major investments in the space – and developed some synthetic video that’s nearly indistinguishable from real footage. We’re seeing hints of a future where not just Netflix or ESPN, but anyone can produce high-quality content, for little cost.
I believe that has major implications for sports. The sports industry is built around selling high-quality content. And as content production fundamentally changes, I’m excited about several opportunities at the intersection of sports and AI video.
I’ll break down this essay into four sections:
A quick primer on the state of AI-generated video
Why AI video matters for the sports industry
Why the sports industry matters for AI video
Where I see opportunities at the intersection of sports and AI video
1. A Quick Primer on AI-Generated Video
Before going deeper, it’s helpful to review the state of AI video.
It all started with Will Smith eating spaghetti. In early 2023, a Redditor shared a video of Will Smith eating spaghetti, made with an open-source Google model. It went viral because it was hilarious, but also somewhat realistic.
From roughly that date, a massive wave of innovation kicked off. Top-tier VCs dumped funds into AI video startups, such as RunwayML ($237M raised), Pika Labs ($135M), and Luma ($73M). Meanwhile, giants like Open AI and Google built out their own research teams, computing resources and training datasets.
AI videos started getting better, fast. The videos became higher-resolution, smoother across frames, and capable of longer durations. Then in March, Open AI shocked the world when it revealed its first video model, Sora. Its videos were fully lifelike. For the first time, the masses grappled with the idea that soon, anyone might be able to create realistic video with just a text prompt.
Open AI’s Sora release video
Today, that future isn’t quite here. AI video models still face many challenges. First, today’s models mostly fail to achieve lifelike quality, especially struggling to nail human faces and bodies. Second, today’s models offer little customization. The same prompt could generate two totally different videos, so it’s challenging to tailor videos to a specific creative vision. Third, computational costs remain high, limiting any mass-market product. Fourth, legal concerns around IP rights and ethical concerns around deep fakes hang over every company in this space. (More on this later).
Despite these roadblocks, AI video quality continues to improve. Here’s a 2024 generation of Will Smith eating spaghetti, compared to just the year prior. If quality continues improving at this rate, we can imagine a future where anyone can create any video content with just a prompt.
2. Why AI Video Matters for the Sports Industry
The future of content creation is consequential – because the sports industry is built on video content.
Starting in the 1970s, cable TV transformed professional sports from a local events business into a massive international media business. Today, the NFL drives 67 percent of its revenue from media deals. The NBA and MLB are both at 49 percent. The bulk of those revenues surround live games rights.
Even in a future with unlimited AI content generation, live games aren’t going anywhere. People care about who wins the Super Bowl or the World Cup because it drives cultural conversation. Just like people get to the office Monday morning and talk about the Presidential election or House of Dragons, people connect over live sports. That’s why live game rights continue to skyrocket in value (see the NFL’s $100B and NBA’s $76B media deals). Live game content is the bedrock of the sports industry – and AI won’t change that.
However, non-live game content is increasingly essential for sports properties. A strong social media has become table stakes for engaging fans, as 9 in 10 Gen Z fans now use social media to consume sports. Companies like WSC Sports and Greenfly have emerged to service teams’ and leagues’ social strategies. Streaming services have also loaded up on non-live sports content. Netflix has now produced a Drive To Survive equivalent with the NFL, NBA, ATP, PGA and multiple other properties – in addition to buying more one-off documentaries like The Last Dance. Furthermore, leagues are building out their content catalogs in order to drive fans to their over-the-top / direct-to-consumer offerings.
Eventually, AI video could service this growing demand for non-live content. Here are just some examples of content series would easily drive viewership across social, streaming and DTC platforms:
A sit-down interview between Yankees Babe Ruth and Derek Jeter
A studio show between the late John Madden and the Manning brothers
Highlights from Michael Jordan’s Dream Team playing LeBron James’ 2024 Olympic team
This content isn’t yet technologically feasible. But AI content is already starting to pop up on social media. For example, Magic Hour AI’s Instagram takes NBA highlights and stylizes them as superheroes, anime and historical figures.
Beyond generating more engaging content, AI video could enable a whole new level of interactivity. Watching sports content could transform from passive to active entertainment. Here are some examples of interactive fan experiences that leagues and teams could sell:
A Zoom call with a teenage Lionel Messi to talk about his life at FC Barcelona’s youth academy
A custom birthday message from Patrick Mahomes based on details you include
A highlight reel showing a superimposed fan playing in an NBA game
In fact, the last example was posted by Overtime last week. I would pay for a full-fidelity version of myself playing in an NBA game.
The key question with all this content is: Who owns the underlying IP, and how is it attributed? If I generate a custom birthday message video from Patrick Mahomes, do the NFL, Chiefs Mahomes, or all the above get compensated, and how much? The answers to these questions are up in the air. But it is clear that if AI video continues to progress, sports properties will have another major content form to monetize their IP.
3. Why the Sports Industry Matters for AI Video
Just as AI video matters for the sports industry, I believe the sports industry matters for AI video.
No AI company has determined the long-term business model for generative video. RunwayML sells a B2B SaaS tool to film studios, ad agencies, and other content producers. Pika Labs sells a B2C creation tool for consumers to make fun videos. Open AI has not released Sora and continues shopping the tech to Hollywood studios and others in search of product-market fit.
Furthermore, it’s unclear what’s the most valuable application for generative video. Perhaps the end-market is entertainment. Perhaps it’s advertising. Either way, I believe sports matter.
If AI video’s end-market is entertainment, sports is the single most in-demand category of entertainment content. In 2023, NFL and college football games accounted for 97 of the top 100 broadcasts in the US. Only the State of the Union, Thanksgiving Day Parade and Oscars cracked the top 100.
Source: Sportico
If AI video’s end-market is advertising, sports have a similarly dominant influence. The world’s largest brands spend billions of dollars every year to create commercials and co-branded content with the top leagues, teams and athletes.
For these reasons, the AI video industry’s winners will have to service the sports industry – by ingesting sports-specific datasets and building models that capture the nuances of sports content. Current models struggle with human faces and bodies, which are obviously critical for sports. But AI video companies will be incentivized to fix these issues in order to tap into the massive TAM of the sports industry.
4. What Opportunities Excite Me the Most
With all that said, I want to share some of the specific opportunities I’m most excited about.
B2B creation & editing tools: RunwayML has grown to a $1.5B valuation by building the AI video creation and editing suite for Hollywood producers and ad agencies. A24 used their tools to create the VFX for Oscar-winning “Everything Everywhere All At Once.” There’s white space for an AI video suite for sports leagues, teams and networks. These tools would cater to sports’ specific content forms (e.g. highlights, graphics) and content pipelines (e.g. real-time editing of nightly game content).
B2C fan engagement experiences: The fan engagement opportunities span every part of the fan experience. Fans could generate highlights combining their all-time favorite legends; create content of them playing in the big leagues for their social media feeds; debate betting picks with AI versions of their favorite commentators; do a meet-and-greet with AI versions of their favorite players. The fan engagement experiences are only bound by our imagination and by the progress of this technology.
New IP creation: When the barriers to creation get lower, new types of sports IP can emerge. In 2021, at the height of the crypto bubble, a group called the Rumble Kong League appeared, promising the next great basketball video game, revolving around their NFT collection. Stephen Curry and Paul George changed their profile pictures to their Rumble Kongz. This showed an insatiable demand for new sports IP. The same is clear from new leagues emerging, from Tiger Woods’ TGL to Shaun White’s S4 (both Will Ventures investments). As content creation becomes easier, someone will create the next Space Jam or Kuroko Basketball IP – or perhaps the next media company based on multiple pieces of IP.
Data labeling & management: Training AI video models requires significant amounts of labeled data. Numerous sport startups from Infinite Athlete to Sportradar already have sophisticated labeling tools to track gameplay and analyze performance. As startups build not just discriminative, but also generative AI models, labeling and managing this data will grow even more important.
Rights & IP management: As AI content proliferates, there will be a need to attribute content and compensate the underlying IP holders. Will Ventures’ strategic partner, OneTeam Partners, currently serves the sponsorship space, bundling players associations’ rights and commercializing them. Genius Sports and Sportradar serve a similar function for the betting industry. I wonder what technologies will be required to accurately attribute IP usage in AI content, and which companies will emerge as the industry’s rights managers.
To summarize, AI video is advancing rapidly. No one knows if or when the technology will hit a wall. But it’s clear that AI video matters for the sports industry, and vice versa – and I’m excited to track this space.
If you’re building at the intersection of sports, entertainment and AI video, please reach out! I’d love to discuss with you.
Amazing post! Love seeing the intersection of AI/LLM and my other favorite rabbit hole to dive into, sports! I sense great things in store for this newsletter. Thanks for sharing!