Scaling Fast: AI’s Boom, Its Bottlenecks, and Its Blind Spots

Week 4: Dec. 15 – 19

AI Unpacked

Welcome to this week’s edition of AI Unpacked, where we break down the most significant developments in artificial intelligence and what they signal for markets, businesses, and policy. This week’s updates include OpenAI’s massive new fundraising round, Meta’s push into AI-generated video, Anthropic’s project Vend, OpenAI’s expanding role in consumer commerce, and several other key updates shaping the industry.

OpenAI’s Fundraising Round Signals Unprecedented Scale

OpenAI is reportedly in discussions for a new fundraising round that could value the company at up to $830 billion, placing it among the most valuable private companies in the world. While the round reflects extraordinary investor confidence in OpenAI’s role as a foundational provider of AI infrastructure, software, and services, it has also reignited debate over whether the AI sector is entering bubble territory.

Such a valuation underscores how rapidly AI has evolved from experimental technology into core economic infrastructure. Investors are betting that OpenAI’s models will power enterprise software, consumer applications, and automation at global scale. However, critics warn that capital is being deployed faster than sustainable revenue can materialize, drawing comparisons to past technology bubbles where expectations outpaced near-term fundamentals.

If completed, the round would further cement OpenAI’s influence over the AI ecosystem, but it would also raise the stakes for proving that today’s massive valuations can be justified by durable business models, long-term cash flows, and real productivity gains rather than speculative momentum.

Meta Develops “Mango,” Its Next AI Image and Video Model

Meta is developing a new AI image and video generation model internally referred to as “Mango,” designed to improve realism, motion consistency, and advanced editing capabilities across images and short-form video. Alongside Mango, Meta is also investing in “Avocado,” a related internal model focused on improving fine-grained visual control, style consistency, and prompt accuracy for creators. Together, these efforts support Meta’s broader strategy to embed generative AI directly into Instagram, Facebook, and future creator tools.

Meta's New AI Image Model "Mango" Source: Nextgenbulletin

The move highlights intensifying competition in AI-generated media, led by OpenAI with Sora, Google with Gemini-powered video tools, and Runway with Gen-3. As these companies race to build more powerful image and video models, the focus is shifting away from novelty toward professional-grade output that can be deployed in advertising, entertainment, and large-scale content production. Meta’s continued investment signals that AI-generated video is quickly becoming a standard tool for creators and marketers rather than an experimental add-on.

Anthropic’s Project Vend Reveals the Limits of AI Agents

Anthropic’s recent Project Vend experiment offered a rare real-world test of how autonomous AI agents perform when given operational control. In the project, Anthropic allowed its Claude model to manage a small vending machine business, giving it authority to monitor inventory, set prices, negotiate restocking, and make purchasing decisions with minimal human oversight.

The results were eye-opening. Over the course of the experiment, Claude made a series of poor business decisions that reportedly led to losses totaling thousands of dollars. The AI frequently overpaid for inventory, failed to adjust prices appropriately in response to demand, and made inefficient restocking choices. In some cases, Claude prioritized politeness or user satisfaction over profitability, highlighting a fundamental misalignment between human-friendly behavior and sound economic decision-making.

Anthropic's Claude-led Vending Machine. Source: Anthropic

While the experiment exposed clear weaknesses, it also served as a valuable stress test for AI agents operating outside controlled environments. Project Vend demonstrated that although AI agents can execute tasks autonomously, they still struggle with long-term strategy, cost control, and real-world incentives. The findings suggest that truly autonomous AI-driven businesses remain far off, and that human oversight is still essential as companies experiment with AI agents in operational roles.

DoorDash and OpenAI Expand AI Into Everyday Commerce

DoorDash has partnered with OpenAI to allow users to shop for groceries directly through ChatGPT. Customers can search for items, compare options, and place grocery orders through natural language prompts instead of traditional apps or websites.

This partnership reflects a broader shift toward conversational commerce, where AI acts as an intermediary between consumers and services. By embedding shopping into AI assistants, companies aim to reduce friction and increase convenience. If successful, this model could significantly alter how consumers interact with retail platforms.

TikTok’s U.S. Joint Venture and the Legal Workaround Debate

Rather than a full sale or shutdown, TikTok’s path forward in the United States is now centered on a newly formed U.S.-based entity called TikTok USDS Joint Venture LLC. Under this structure, TikTok’s American operations would be housed within a separate domestic company designed to satisfy U.S. regulatory requirements while allowing the platform to continue operating without major disruption.

According to an internal memo from CEO Shou Zi Chew, TikTok USDS Joint Venture LLC has signed formal agreements with three managing investors: Oracle, Silver Lake, and MGX. Oracle is expected to play a central role in data storage and security, while Silver Lake and MGX provide financial backing and governance participation. Together, the structure is meant to demonstrate U.S. oversight, domestic control, and compliance with national security concerns.

Critics argue, however, that this joint venture effectively bypasses the spirit of earlier U.S. mandates aimed at limiting Chinese influence over TikTok. While the entity may meet the technical requirements of U.S. law, skeptics warn that key elements, such as recommendation algorithms, intellectual property, and strategic decision-making, could remain indirectly tied to TikTok’s China-based parent company. From this perspective, the arrangement functions less as a true divestment and more as a legal workaround that preserves TikTok’s U.S. business.

Supporters counter that the joint venture strikes a pragmatic balance, protecting millions of users, creators, and advertisers while introducing enforceable safeguards around data access and governance. The debate underscores a broader challenge for regulators: as AI-driven platforms grow more complex, ownership structures can be reshaped faster than laws can adapt. TikTok’s U.S. joint venture may ultimately serve as a precedent for how foreign AI platforms navigate regulation without fully relinquishing control.

House Passes Permit Reform Bill to Accelerate AI Infrastructure

The U.S. House of Representatives has approved the SPEED Act, a permitting reform bill designed to accelerate approval timelines for infrastructure critical to artificial intelligence, including data centers, semiconductor facilities, and energy projects. Supporters argue that faster permitting is essential to meet surging AI-related power demand and to maintain U.S. competitiveness as global investment in AI infrastructure accelerates.

The bill has gained backing from major technology companies such as OpenAI, Micron, and Microsoft, which have all emphasized that regulatory delays are becoming a bottleneck for large-scale AI deployment. While the House has moved forward, the Senate remains in the early stages of its own permitting discussions, leaving the final scope and timing of reform uncertain.

As AI models require enormous computing and energy resources, infrastructure constraints are increasingly shaping the pace of innovation. The SPEED Act signals growing recognition in Washington that leadership in AI depends not only on software breakthroughs, but also on energy availability, construction speed, and long-term infrastructure planning.

Looking Ahead

This week’s developments show AI expanding simultaneously across finance, media, commerce, infrastructure, and geopolitics. From record-breaking valuations to autonomous agents and regulatory reform, the technology is moving deeper into the core of the economy. As AI systems grow more capable and more embedded in daily life, the balance between innovation, oversight, and scale will define the next phase of the industry.

Stay tuned for next week’s breakdown as we continue tracking how AI is reshaping business and technology in real time.