A quiet revelation is emerging: Waymo’s robotaxis may drive themselves on U.S. roads, but many of their most critical decisions are still made by human operators sitting thousands of miles away in the Philippines. Waymo, after all, is owned by Google’s parent company. Autonomy, it seems, is layered.

Waymo has long been held up as the gold standard of autonomous driving, a symbol of how far artificial intelligence has progressed beyond human control. Yet testimony at a recent U.S. Senate hearing has exposed a less visible but increasingly important reality: when Waymo’s vehicles encounter situations they cannot resolve on their own, responsibility often shifts to remote human workers, many of whom are based in the Philippines.

The disclosure by Waymo’s chief safety officer, Mauricio Peña, cut through years of marketing around “driverless” technology. Peña told lawmakers that in rare or complex scenarios, Waymo’s robotaxis can hand over control to remote operators who guide the vehicle through the situation. These workers act as a form of last-resort intelligence, stepping in when sensors, software, and pre-trained models are insufficient to safely navigate the real world.

This reminds me of an experience many of us had during industrial training in Nigeria’s oil and gas industry. You would leave Port Harcourt in a new vehicle, driven by a polished professional. Everything felt regulated, certified, and safe. Then, somewhere around Ogoni on the way to Ikot Abasi, the road ended, and the sea began.

To cross it, you boarded a small speedboat manned by a driver who might not even be allowed through the gates of the oil company, often visibly drinking. Once the boat took off, all power and influence shifted. The man at the helm was a human “fish”, someone who could survive what you could not in the sea. Ironically, most people on board had swimming certificates, yet none could swim well enough to matter. In that moment, safety protocols dissolved, replaced by an uncomfortable trust in informal expertise. Yes, the roads are great for professional drivers but on the sea, anything could go!

There is an Igbo saying: a bird that flies off the ground only to perch on an anthill is still very much on the ground. Waymo’s autonomy fits this wisdom. If its most critical decisions are still made by human operators in the Philippines, then the real question is no longer just about Waymo’s engineering.

It is about who those humans are, how they are trained, and how much judgment we are delegating to people we never see. In the age of AI, technology may scale but accountability remains stubbornly human.

Waymo’s robotaxis may drive themselves on U.S. roads, but many of their most critical decisions are still made by human operators sitting thousands of miles away in the Philippines.

Waymo has long been held up as the gold standard of autonomous driving, a symbol of how far artificial intelligence has progressed beyond human control. Yet testimony at a recent U.S. Senate hearing has exposed a less visible but increasingly important reality: when Waymo’s vehicles encounter situations they cannot resolve on their own, responsibility often shifts to remote human workers, many of whom are based in the Philippines.

The disclosure by Waymo’s chief safety officer, Mauricio Peña, cut through years of marketing around “driverless” technology. Peña told lawmakers that in rare or complex scenarios, Waymo’s robotaxis can hand over control to remote operators who guide the vehicle through the situation. These workers act as a form of last-resort intelligence, stepping in when sensors, software, and pre-trained models are insufficient to safely navigate the real world.

What unsettled lawmakers was not simply the existence of human intervention, but where that intervention takes place. The Philippines has become a global hub for outsourced digital labor, from call centers to content moderation and data labeling. Waymo’s reliance on Filipino contractors places the country at the center of America’s most advanced autonomous driving program, even as public messaging continues to emphasize full autonomy.

The logic is partly economic and partly structural for Waymo. Remote driving and support roles require a large, always-available workforce trained to respond quickly to unpredictable scenarios. The Philippines offers a deep labor pool with strong English proficiency and long experience supporting U.S. technology firms. Costs are lower than in the United States, and maintaining round-the-clock coverage across time zones is easier. In practice, this makes the Philippine workforce a quiet but critical component of Waymo’s safety architecture.

The arrangement, however, raises uncomfortable questions about what autonomy really means. If a robotaxi in San Francisco freezes at a construction zone or behaves unpredictably around emergency vehicles, a human operator in Manila or Cebu may be the one deciding how it proceeds. That human judgment, mediated through screens and networks, becomes part of the driving system itself. Autonomy, in this sense, is not the absence of humans but a reorganization of where and how their labor is used.

But this has prompted safety concerns. Senators pressed Peña on latency and reliability, given the physical distance between vehicles and operators. Even small delays in communication could matter in traffic situations unfolding in seconds. Peña maintained that Waymo has built safeguards into its systems and that remote intervention is tightly controlled. Still, the hearing underscored a basic tension: the more robotaxis are deployed at scale, the more edge cases arise, and the more human backup is required.

The focus on foreign workers also reflects a broader shift in Washington’s thinking about technology and national control. Massachusetts Senator Ed Markey called the use of overseas remote drivers “completely unacceptable,” framing the issue not just as a labor question but as one of sovereignty and security. Lawmakers voiced unease about critical transportation decisions being influenced by workers outside the United States, particularly as autonomous vehicles become more integrated into urban infrastructure.

Waymo’s case is especially sensitive because of its hardware choices. Unlike Tesla, which builds and controls its own vehicles, Waymo uses cars manufactured in multiple countries, including China. Although Peña emphasized that Waymo’s autonomous systems are installed and managed in the U.S., the combination of foreign-built vehicles and foreign-based operators has fueled suspicions about vulnerabilities in the system. In an era of heightened scrutiny over supply chains and data flows, even indirect links to China or other overseas networks attract political attention.

For the Philippine workforce itself, the role highlights another recurring pattern in the AI economy: essential labor that remains largely invisible. Much like the content moderators and data annotators who helped train large language models, remote operators supporting robotaxis occupy a gray zone. They are central to system performance but rarely acknowledged in public narratives about innovation. Pay, working conditions, and long-term career prospects for these workers are seldom discussed, even as their decisions can carry real-world consequences.

Although Waymo is not alone in this model, its prominence makes it a test case. The company has positioned itself as a leader in safe, scalable autonomy, operating commercial robotaxi services in multiple U.S. cities. As deployments expand, reliance on remote human support may grow rather than shrink, at least in the near term. That reality complicates claims that autonomy will soon eliminate human involvement in driving altogether.

The Senate hearing suggests that regulators are beginning to look past the surface of AI systems to examine the labor structures beneath them. Scrutiny of Waymo’s Philippine workforce is unlikely to fade. It touches on safety, labor practices, national security, and the credibility of autonomy itself. The technology may be cutting-edge, but its foundations rest on human judgment — relocated, outsourced, and largely unseen.

Alphabet’s AI spending spree is reshaping its balance sheet, risk profile, and core business model, marking a turning point in how even Big Tech funds and manages growth.

Alphabet is turning again to the debt market to help bankroll one of the most aggressive artificial intelligence investment drives in corporate history, while simultaneously flagging fresh risks tied to the technology’s rapid rise and the sheer scale of its infrastructure buildout.

The renewed trip to the debt market is seen as a signal that the artificial intelligence boom has reached a scale where even one of the world’s most cash-generative companies is rethinking how it finances growth — and openly acknowledging the risks that come with it.

In its latest annual filing, released alongside earnings, the Google parent laid out a more cautious narrative around AI than investors have been accustomed to hearing. While executives continue to frame AI as a once-in-a-generation opportunity, the company also warned that the speed, cost, and uncertainty of the buildout could strain operations, expose it to financial liabilities, and, in a worst-case scenario, leave it sitting on underutilized infrastructure.

That backdrop explains why Alphabet is preparing to raise about $20 billion through a multi-tranche bond sale, including an ultra-long 100-year sterling bond, according to people familiar with the deal quoted by CNBC. Demand has been strong, with the offering reportedly several times oversubscribed, underscoring how eager investors remain to lend to top-tier technology names even as borrowing rises.

The fundraising follows a $25 billion bond sale late last year and caps a sharp pivot in Alphabet’s capital structure. Long-term debt quadrupled in 2025 to $46.5 billion, a striking shift for a company that, for years, relied overwhelmingly on internal cash flows. The reason is simple: the scale of AI investment now dwarfs what operating cash alone can comfortably support without trade-offs.

Alphabet said it may spend up to $185 billion in capital expenditures this year, more than double its 2025 outlay. That figure places it firmly among the biggest corporate spenders on infrastructure in history. The money is being poured into data centers, specialized chips, custom silicon, power generation, cooling systems, and high-capacity networking — all essential to training and running large language models such as Gemini.

Yet the filing makes clear that this is not just about owning servers. Alphabet is increasingly relying on long-term leasing arrangements with third-party data-center operators to secure capacity quickly. While this approach accelerates deployment, it also raises costs and complexity, and creates contractual obligations that could become a burden if demand forecasts miss the mark.

Large commercial agreements, the company warned, could increase liabilities if Alphabet or its partners fail to perform. This is a notable admission in an industry that has largely emphasized upside while downplaying execution risk.

Management is keenly aware of the tension. Chief financial officer Anat Ashkenazi told analysts the company wants to invest “in a fiscally responsible way” and preserve a strong financial position. But when CEO Sundar Pichai was asked what worries him most, his answer was blunt: compute capacity. Power availability, land, supply chains, and the pace of expansion now dominate the strategic agenda.

Alphabet’s predicament mirrors a broader shift across Big Tech. Microsoft, Meta, and Amazon are all dramatically increasing capital spending, and together with Alphabet are projected to lift capex by more than 60% this year compared with already record levels in 2025. The collective outlay is fueling what Nvidia CEO Jensen Huang has described as the largest infrastructure buildout in human history.

For Alphabet, however, the stakes extend beyond infrastructure efficiency. AI is beginning to intersect directly with its core advertising business, which still accounts for the majority of revenue and profits. As generative AI tools become more capable, there is a growing question about how users interact with the internet — and whether traditional search, the backbone of Google’s ad machine, could be disrupted.

That concern appeared explicitly in Alphabet’s risk disclosures for the first time. The company acknowledged that consumers may reduce their use of conventional search as AI assistants answer questions directly, potentially altering traffic patterns and advertising formats. Alphabet said it is adapting with new ad products and strategies, but cautioned there is no guarantee these efforts will succeed.

So far, financial performance suggests resilience. Advertising revenue rose 13.5% year on year in the fourth quarter to $82.28 billion, easing fears that AI is already eroding demand. But investors are looking beyond current results to the medium-term implications of a shift in user behavior that could be structural rather than cyclical.

At the center of Alphabet’s AI push is Gemini, its flagship model and assistant, which now boasts more than 750 million monthly active users, up sharply from the previous quarter. The rapid uptake underscores why the company feels compelled to invest at scale: falling behind rivals such as OpenAI or Anthropic would risk ceding influence over the next generation of computing interfaces.

Yet scale cuts both ways. The more capital Alphabet commits upfront, the greater the risk of overcapacity if monetization lags, competition intensifies, or technological change renders today’s infrastructure less valuable. The company’s own warning about “excess capacity” reflects a growing recognition that the AI arms race may not deliver returns evenly or immediately.

Alphabet’s decision to lean more heavily on debt also reflects a subtle recalibration of financial strategy. Borrowing allows the company to preserve cash for flexibility, smooth out investment cycles, and take advantage of still-favorable credit markets. But it also ties Alphabet more closely to investor sentiment and interest-rate dynamics, at a time when capital markets are increasingly sensitive to execution risk.

In that sense, the bond sale is symbolic. It shows how AI has pushed Big Tech into territory once associated with capital-intensive industries such as energy or telecommunications, where long-dated assets, fixed costs, and leverage are part of the business model.

Currently, Alphabet retains enormous advantages: scale, data, engineering talent, and a balance sheet that remains strong by almost any measure. But its latest filings and financing plans suggest a more candid tone about the challenges ahead. The AI boom may still be in its early innings, but Alphabet is already grappling with the reality that building the future of computing is expensive, complex, and fraught with trade-offs that even Google cannot fully control.