AWS wants you to believe it’s become indispensable to entertainment. Samira Bakhtiar, who runs the cloud giant’s global media, entertainment, games and sports division, spent a December afternoon at AWS re:Invent making that case to a handful of journalists. Her pitch wasn’t subtle: 750 million gamers use AWS-powered services monthly, studios are experimenting with AI-generated content at scale, and the infrastructure tying all of this together runs on Amazon’s servers.
It’s an aggressive positioning play, and it’s working. AWS has quietly embedded itself across Hollywood’s post-production workflows, gaming infrastructure, and now, increasingly, the AI tools that studios are testing for everything from storyboarding to deepfake detection. The company isn’t just providing cloud storage anymore. It’s building the rails for what Bakhtiar describes as entertainment’s “inflection point”, where content creation speed matters as much as creative vision.
Sony’s on-premise AI experiment signals Hollywood’s nervous approach
The most telling example Bakhtiar shared wasn’t about cloud services at all. Sony Pictures is using Nova Forge, an AWS service announced at re:Invent, to train foundational AI models on-premise with proprietary data. That’s the opposite of cloud-first thinking. It’s studios hedging their bets, wanting AI’s capabilities without surrendering control of decades of intellectual property to someone else’s data centre.
John Kidera, a Sony executive, discussed the approach during the keynote Bakhtiar referenced. The use case centred on Demon Slayer, an anime property Sony has been monetising across formats. The implication is clear: studios want AI to assist with production workflows, personalised streaming experiences, and content distribution, but they’re not ready to hand over their crown jewels. AWS is accommodating that caution, even if it means compromising on the cloud orthodoxy it’s spent years evangelising.
Startups are filling Hollywood’s AI skills gap faster than studios can hire
Bakhtiar spent considerable time discussing the startup ecosystem building AI tools for media production on AWS infrastructure. Luma AI, Crea.ai, and Loti.ai were named repeatedly. These aren’t household names, but they’re building the scaffolding Hollywood needs to experiment with synthetic content creation, automated post-production, and deepfake monitoring.
Loti.ai’s service particularly highlights the dual-edged nature of AI in entertainment. Users can upload an image and the platform crawls the web to detect unauthorised deepfakes. It’s a reactive solution to a problem the industry helped create by rushing towards synthetic media without adequate safeguards. Bakhtiar framed this as studios “meeting the moment of reinvention”, which is generous. It’s more accurate to say they’re scrambling to manage risks they didn’t anticipate.
The reliance on startups also exposes a gap. Established studios don’t have the internal expertise to build these tools themselves, so they’re outsourcing to companies that might not exist in five years. AWS benefits either way because the infrastructure layer remains constant, but it’s a precarious model for content creators betting their production timelines on early-stage companies.
Fortnite keynotes and in-game concerts blur the lines between platforms
AWS streamed its re:Invent keynote inside Fortnite this year, which Bakhtiar described as a “full-circle moment”. The technical execution relied on AWS Elemental Media Services, traditionally used for live sports and news broadcasts, now repurposed for in-game distribution. Epic Games, which owns Fortnite, runs on AWS infrastructure, so the stunt doubled as both a customer showcase and a marketing exercise.
The broader trend Bakhtiar is pointing towards is convergence. Entertainment companies are testing distribution models that don’t rely on traditional streaming platforms or broadcast schedules. Fortnite and Roblox have both hosted virtual concerts, pulling tens of millions of concurrent viewers. IMAX, a company built on premium theatre experiences, streamed NBA finals games to reach younger audiences who don’t watch sports on television.
This isn’t new behaviour, but the speed at which companies are experimenting has accelerated. Bakhtiar attributed this to AWS’s combination of purpose-built services, scalable compute, and AI tools reaching maturity simultaneously. That’s partially true, but it’s also driven by desperation. Traditional media companies are haemorrhaging younger audiences to gaming, social platforms, and short-form content. They’re testing anything that might reverse the decline, and AWS is happy to provide the infrastructure.
Agentic AI promises operational savings, but at what cost?
Bakhtiar outlined three areas where agentic AI will reshape media operations: automating video ingestion and metadata tagging, enabling real-time ad placement in live news, and streamlining media planning and bidding processes. These are legitimate efficiency gains. Media operations centres currently employ dozens of people watching screens to manually tag content, ensure quality control, and detect ad breaks. Autonomous agents can handle most of that work.
The trade-off is employment. Bakhtiar framed automation as freeing workers for “creative innovations”, which is the standard line every company uses when explaining why roles are being eliminated. Media operations jobs aren’t glamorous, but they’re stable and they employ people who understand the nuances of content production. Replacing them with AI agents that require constant monitoring and fine-tuning doesn’t necessarily improve output quality. It just shifts costs.
The advertising use case is more compelling. News outlets struggle to monetise effectively because brands don’t want their ads appearing next to coverage of disasters, political scandals, or violent crime. Agentic AI can analyse live feeds in real-time, categorise content appropriateness, and match inventory dynamically. That’s a tangible revenue opportunity, assuming the systems work reliably and don’t make embarrassing placement errors that damage brand relationships.
Content archives are the real prize, and AWS wants to index them
One of Bakhtiar’s repeated themes was the untapped value in entertainment companies’ archival content. Decades of film, television, and recorded material sit in warehouses or on legacy storage systems, inaccessible because there’s no searchable metadata. AWS’s pitch is straightforward: ingest everything into the cloud, use AI to extract metadata, and suddenly that content becomes monetisable again.
BBC and Banijay were cited as examples of companies using AWS to process hundreds of thousands of hours of archival material. The output enables personalised streaming recommendations, short-form content for social platforms, and licensing opportunities previously hidden in unsearchable libraries. [AWS’s investment in South African cloud infrastructure](https://reframed.co/first-international-aws-skills-center-opens-in-cape-town/) suggests similar strategies could apply to broadcasters and studios across Africa with deep content archives.
The technical capability exists. AWS launched S3 vector embedding for video understanding, partnered with 12Labs on the Marengo 3 model for deep video analysis, and embedded search functionality into Nova models. The challenge isn’t technology. It’s rights management. Much of that archival content involves complex licensing agreements, union restrictions, and talent contracts that didn’t anticipate AI-driven reuse. Studios will spend years navigating legal frameworks before they can fully exploit these capabilities.
Gaming infrastructure quietly powers AWS’s entertainment credibility
Bakhtiar’s claim that 750 million gamers use AWS-powered services monthly is difficult to verify, but the customer list is credible. Epic Games, Electronic Arts, Roblox, Scopely, Zynga, Take-Two, and even The New York Times for Wordle all run game infrastructure on AWS. Services like GameLift and GameLift Streams handle the hyperscale demands of multiplayer games, which require ultra-low latency and high availability.
The DraftKings example Bakhtiar mentioned highlights how gaming infrastructure translates to adjacent markets. DraftKings, a sports betting platform, deployed a streaming service within its app using AWS to let users watch games while placing live bets. The regulatory requirements for betting platforms, particularly around latency, security, and compliance, mirror the technical demands of gaming. AWS is positioning itself as the infrastructure layer that can handle both.
The company’s betting and gaming summit at re:Invent attracted 140 customers from 33 countries, which suggests there’s significant demand for regulated gambling infrastructure. That’s not surprising. Betting operators need systems that can process transactions in milliseconds, maintain audit trails for regulators, and scale during major sporting events. AWS’s advantage is that it already built those capabilities for gaming and can repackage them for a different vertical.
Personalisation demands are outpacing content libraries
Bakhtiar made an observation worth sitting with: attention spans are shrinking while demand for personalised, immersive content is intensifying. That’s a brutal dynamic for entertainment companies. Audiences expect recommendations tailored to their specific preferences, not just demographic categories, but content catalogues aren’t growing fast enough to satisfy that demand.
The solution Bakhtiar described involves using AI to extract more value from existing content. Personalised streaming experiences, dynamic ad insertion, short-form social clips, and influencer licensing partnerships all aim to make individual pieces of content work harder. It’s a scarcity mindset dressed up as innovation.
There’s also a ceiling to personalisation. At some point, hyper-targeted recommendations start to feel intrusive rather than helpful. Streaming platforms already face criticism for filter bubbles and algorithmic curation that limits discovery. Doubling down on personalisation might improve engagement metrics in the short term, but it risks making the content experience more isolating.
AWS isn’t neutral infrastructure anymore
What emerged from Bakhtiar’s presentation is that AWS has moved beyond being a passive infrastructure provider. The company is actively shaping how entertainment is produced, distributed, and monetised. That’s not inherently problematic, but it does concentrate power. When the same company provides cloud storage, AI training tools, content distribution services, and gaming infrastructure, it becomes difficult for customers to switch providers or negotiate from a position of strength.
Bakhtiar’s description of AWS as “the backbone of entertainment” isn’t hyperbole. It’s strategy. The more deeply embedded AWS becomes in production workflows, the harder it is for studios, broadcasters, and gaming companies to move to competitors. Lock-in doesn’t happen through contracts. It happens through dependency.
The entertainment industry is experimenting with AI at a pace that would’ve seemed reckless five years ago. Studios are testing synthetic content tools, automating post-production workflows, and building entirely new distribution models inside gaming platforms. AWS is providing the infrastructure that makes those experiments possible, but it’s also extracting value at every layer. Hollywood might be driving the creative vision, but Amazon is building the roads.