Veo 3.1 Lite: High-Volume AI Video at Half the Cost
If you build anything with video generation, or have been watching the cost of AI video pipelines with concern, Veo 3.1 Lite is the most directly relevant release of the day. Announced via Google's official blog and rolled out immediately to the Gemini API and Google AI Studio, the new model generates video at less than 50 per cent of the cost of Veo 3.1 Fast while maintaining the same generation speed. For developers running high-volume applications, that is a significant change in unit economics.
Veo 3.1 Lite supports Text-to-Video and Image-to-Video, with outputs available at 720p or 1080p and in both landscape (16:9) and portrait (9:16) aspect ratios. Clip duration is flexible at 4, 6, or 8 seconds, with cost scaling accordingly. The launch signals that the Veo 3.1 model family is now complete, with three tiers (Pro, Fast, Lite) covering quality-critical, balanced, and cost-optimised workloads respectively. Google also flagged that pricing for Veo 3.1 Fast will be reduced further from 7 April, compressing costs across the entire family.
The timing is notable: it lands in the same news cycle as continuing coverage of OpenAI's decision to wind down Sora (announced 24 March, with the app closing 26 April). Where OpenAI treated video generation as a creative showcase that struggled to find a commercial model, Google is positioning Veo 3.1 Lite squarely at developers and API consumers who need to run video at scale without budget-breaking per-clip costs. The competitive gap in developer-ready video generation has widened considerably today.
OpenAI Closes $122 Billion Funding Round at $852 Billion Valuation
OpenAI has completed the largest private financing round in history, raising $122 billion at an $852 billion post-money valuation. If you use OpenAI's products or are watching the competitive dynamics between the major AI labs, this matters less for what it means today and more for what it signals about the next 18 to 24 months. OpenAI now has near-unlimited runway to expand its compute infrastructure, hire talent, and subsidise usage pricing to win market share in a way that smaller competitors simply cannot match.
Amazon contributed $50 billion, Nvidia and SoftBank each put in $30 billion, while a16z, D.E. Shaw Ventures, MGX, and TPG co-led the remainder. Notably, over $3 billion was raised directly from retail investors through bank channels, a first for the company. Amazon's full commitment of $50 billion is contingent on OpenAI either going public or reaching a stated AGI milestone, two triggers that will be closely watched by the industry.
OpenAI stated the funds will go towards expanded compute infrastructure, multi-cloud partnerships, and building a unified AI superapp that combines ChatGPT, Codex, search, and agentic capabilities into a single product surface. The company is now generating $2 billion in monthly revenue with approximately 900 million weekly active users. For developers evaluating platform dependence, this round makes it harder to bet against OpenAI's longevity but also harder to predict how aggressively they may shift pricing once they have pricing power.
MCP Dev Summit North America Opens in New York City Today
If you are building agentic systems or integrating the Model Context Protocol into your workflows, today is worth paying close attention to. The MCP Dev Summit North America opens in New York City today (2 April) and runs through 3 April, organised by the Agentic AI Foundation (AAIF), the Linux Foundation directed fund launched in December 2025 by Anthropic, OpenAI, and Block to steward open standards for AI agents. The conference is the first major in-person gathering dedicated to MCP in production, and its sessions reflect a protocol that has moved well beyond early-adopter experimentation.
The programme spans more than 95 sessions covering protocol evolution, conformance testing, security research, and real-world deployment patterns. Speakers include engineers from Anthropic, Microsoft, Datadog, and Hugging Face. Notable sessions include a deep-dive into MCP at 18 months (Hugging Face), a session on multi-issuer security attacks in MCP environments (Microsoft), and a look at when MCP is not the right architectural choice for scalable agent systems (Datadog). The emphasis on security and production readiness signals that the MCP ecosystem is entering a maturity phase where robustness matters as much as capability.
Industry Themes
Developer cost efficiency is becoming the dominant competitive axis for AI infrastructure products. Google's Veo 3.1 Lite launch was framed explicitly around per-unit cost reduction for high-volume applications, not capability improvement, and Google flagged further price cuts for Veo 3.1 Fast within days. This mirrors the trajectory of text model pricing over the past two years, where successive model families collapsed per-token costs by orders of magnitude. Video is following the same curve, and today's launch suggests Google is moving deliberately to fill the vacuum left by Sora's exit.
The agentic infrastructure layer is moving into production maturity. The MCP Dev Summit's programme, with its emphasis on conformance testing, security research, and deployment lessons rather than introductory tutorials, reflects a protocol that has moved from interesting to critical for teams building production AI systems. The session roster includes engineers from Anthropic, Microsoft, Datadog, and Hugging Face, which is a strong signal that MCP is now considered a serious enterprise infrastructure component rather than an experiment.
The capital gap between frontier labs and everyone else is widening sharply. OpenAI's $122 billion close gives it resources that few competitors can approach, and its stated plan to consolidate ChatGPT, Codex, and agentic capabilities into a single superapp means the battle for developer attention will intensify on multiple fronts simultaneously. For teams evaluating which foundation model platforms to build on, today's funding news is a signal that OpenAI's product investment is unlikely to slow, but also a prompt to consider what platform concentration risk looks like at the $852 billion scale.