Summary of Key Points
Meta's second-quarter financial report reveals a stark contrast between increased revenue and declining profits: Its AI-driven advertising business saw revenue grow by 28% (to $60.8 billion), but net profit plummeted by 14% (to $15.8 billion), with free cash flow dropping by 91% to just $784 million. Meanwhile, Meta, along with Silicon Valley giants like NVIDIA and Elon Musk, has publicly supported open-source AI, opposing the ban on Chinese AI models. This is not an ideological stance but a practical choice driven by massive hardware investment. Open-source technology can reduce innovation costs and accelerate the adoption of AI, helping them recoup their billions in infrastructure investments more quickly.
The Good and Bad News in the Financial Report: AI Advertising Generates Revenue, but Expenses Are Rising Rapidly
Good News: AI advertising has saved Meta's revenue. Meta's advertising business has seen both increased volume and higher prices due to AI algorithms that make ads more targeted, attracting more advertisers willing to pay higher fees, as well as more user clicks and conversions.
Bad News: Profits and cash flow are under pressure. Net profit decreased by 14% mainly because of high expenses: R&D spending rose by 67% (to $21.7 billion), and capital expenditures (for data centers and AI hardware) soared to $31.08 billion, with annual plans for $130-145 billion. Free cash flow has dropped by 91% to $784 million—similar to a situation where your salary increases, but you spend more on housing and renovations, leaving you with little money left.
The problem is that Meta doesn't have a cloud business like Microsoft Azure or Google Cloud to directly profit from selling computing power. Although some suggested renting out excess computing resources, Zuckerberg rejected the idea, believing that selling computing power would only generate short-term profits compared to using it for intelligent products (such as ad recommendations and AI assistants) with higher margins. However, this requires time, and the current cash flow pressure is significant.
Why Is Meta Urgent to Promote Open-Source AI? To Recoup Its Billion-Dollar Investments
Meta is investing heavily in building superdata centers, such as a $14 billion facility in Texas and a $50 billion Hyperion center in Louisiana, with the goal of maximizing computing capacity by 2026-2027. But how will they recoup this investment?
Open-source AI is crucial. Open-source models (like Meta's Llama) allow more developers to use them for free or at low costs, lowering the barriers to using AI applications. This:
1. Reduces Meta's costs: Lower operating expenses for these models can alleviate profit pressure.
2. Expands the ecosystem: More users of open-source models will rely on Meta's technology, indirectly boosting advertising and other businesses.
3. Accelerates revenue generation: A more prosperous open-source ecosystem means a faster return on Meta's hardware investments.
If the AI ecosystem remains closed and only a few companies (like OpenAI) can use it, Meta's billion-dollar investments could go unrealized.
Why Do Silicon Valley Giants Support Open-Source AI? Their Own Self-Interests Are at Play
Not only Meta but also NVIDIA and Musk support open-source technology for commercial reasons:
- NVIDIA: Jensen Huang says Chinese models are excellent and should be used. He believes that Chinese open-source models will lower the cost of using AI, leading to more purchases of NVIDIA chips (such as the H100). He even suggests that Wall Street underestimates the impact of Chinese models like DeepSeek and Kimi.
- Elon Musk: He has long advocated for open-source AI, fearing that closed models could be monopolized by a few companies. Additionally, Tesla's autonomous driving technology benefits from an open-source ecosystem.
- Closed-Source Companies (OpenAI/Anthropic): They oppose open-source because it threatens their high subscription fees. If everyone uses free open-source models, their businesses would be affected.
The debate over open-source vs. closed-source is essentially about how to share the “cake”: NVIDIA wants to grow the market by selling more chips, Meta wants to profit from the open-source ecosystem, and closed-source companies want to maintain their monopoly.
Meta's Smart Strategy: A Balanced Approach
Zuckerberg states that Meta won't rely solely on open-source models, as they may not be as powerful as closed-source ones like GPT-4. Meta's strategy is a “full-stack approach”:
- Continuing to develop open-source models (Llama series) to maintain a low-cost ecosystem.
- Researching and developing its own cutting-edge closed-source models to stay technologically ahead.
- Building its own data centers and manufacturing chips (like the MTIA chip) to control the entire cost chain.
This balanced approach avoids the risk of falling behind in open-source technology while leveraging its benefits to reduce costs, offering both affordable (open-source) and premium (closed-source) solutions to meet different user needs.
Opposing Bans on Chinese Models: Fear of a Divided Ecosystem and Increased Costs
Zuckerberg firmly opposes U.S. bans on Chinese AI models, arguing that they would split the global AI ecosystem into two parts: “Western closed-source” and “other open-source.” This would limit Meta's ability to reach global markets and extend the return on its billion-dollar investments. The low-cost advantage of Chinese models can also help reduce Meta's operating expenses. Huang Renxun agrees, stating that competition should encourage innovation rather than be restricted by bans, as such measures benefit only closed-source companies like OpenAI at the expense of overall industry progress.
In Conclusion: Giants' Stances Are All About Business
Meta and Silicon Valley giants' support for open-source technology is driven by business interests. They need an open-source ecosystem to offset their heavy hardware investments and a global market to accelerate revenue generation. The future of AI competition will depend on which models are more powerful and whose ecosystems are more prosperous. Current alliances are temporary, and the real battle has just begun.