Which AI Company Is Most at Risk? Chen Fengxin Names One Giant: Stock Plunged 41%, One More Downgrade to 'Junk' Status

Recent financial reports from major U.S. tech companies have triggered market concerns over massive AI-related spending. Senior media personality Chen Fengxin, on her program 'East-West North-South Longfeng Pei,' pointed out that this year, the capital expenditures of five tech firms are expected to exceed the cash flow generated from operations, resulting in a combined negative free cash flow of $2.8 billion. The most at-risk company is Oracle. Oracle's stock has dropped 41% year-to-date, and S&P Global downgraded its bond rating from BBB to BBB-, one level above junk status. A further downgrade would classify it as a 'junk' company.

Why Did Alphabet's Stock Fall? Chen Fengxin explained that last week, Alphabet, Google's parent company, released its Q2 earnings. Revenue and profits were very high, yet the stock plunged 7% after the announcement. The main reason was the upward revision of capital expenditures. In theory, higher capital spending should benefit the supply chain due to surging computing demand, but why did the stock fall?

Chen noted that in the past, Google's capital spending wasn't entirely AI-driven, but since 2022 and 2023, it has rapidly increased due to AI demand. For years, Google had ample cash but lacked investment targets, keeping capital spending low. Now, however, Google's spending has surged, with estimated annual capital expenditures reaching $200 billion. In Q2 alone, capital spending jumped to over $40 billion. Google has burned through its cash reserves, which are now negative $5.9 billion.

Has Google Burned Through Its Cash? Chen Fengxin stated that while $5.9 billion may seem insignificant relative to Google's revenue and profits, the fact that a company with consistently high cash reserves for over a decade has now depleted its cash is alarming. Capital expenditures have not only consumed all profits from core operations like YouTube ads and search engine revenue but have also exhausted existing cash reserves, with further spending planned. The market now questions how long this spending model can be sustained.

Which Mega Cloud Service Company Has Fallen the Most? Chen pointed out that in the first half of this year, stocks that rose were upstream suppliers such as mega cloud data center supply chains, semiconductor firms, memory companies, passive components, and optical communication firms. However, the cloud service companies actually investing heavily in AI infrastructure have mostly declined. As of the 27th, Google rose slightly by 1.45%, Amazon by 2.47%, but the other four companies all fell, with declines ranging from 8.5% to 41% as of the 25th. Oracle fell the most, down 41% year-to-date. Initially, concerns were limited to these six companies, but now doubts are spreading to others.

Chen noted that while these companies are profitable, their capital expenditures are rising so rapidly that this year’s spending is expected to exceed operating cash flow, resulting in a combined negative free cash flow of $2.8 billion. This figure was calculated before Alphabet’s earnings release and may worsen upon further scrutiny.

Is Oracle the Most at Risk? Chen emphasized that Oracle is actually the most dangerous. Its stock has fallen 41% year-to-date. Oracle is the 'canary in the coal mine' for AI infrastructure, so problems will surface here first. Over the past four years, from 2023 to Q4 2024, Oracle has reported negative free cash flow every quarter, accumulating losses. Shouldn’t the situation improve in Q4? Yet, why does Oracle’s stock continue to fall?

Chen explained that when Oracle disclosed these figures in its Q1 earnings, the stock decline didn’t slow. Now, Oracle and many large data centers are adopting the same strategy: hiding liabilities. To build an AI data center requiring $10 billion, do they actually spend $10 billion? No. They require customers who have already contracted for computing power to provide their own GPUs, while the company only builds the physical infrastructure. The most expensive component—the GPU—is paid for by the customer, so the company avoids GPU purchase costs.

Can Tech Companies Hide Liabilities? Chen noted that OpenAI and Anthropic buy and install GPUs themselves. GPUs account for about 60–70% of total costs. By not paying for 60–70%, the company only spends $3–4 billion of the $10 billion. But even that isn’t fully paid upfront. The company then charges customers five years of rent, recovering about 20% of costs. Thus, the capital expenditure recorded on the books shrinks from $10 billion to just $1–2 billion, making cash flow appear healthy.

However, Chen warned that if data center construction proceeds smoothly, this reduction in free cash flow isn’t problematic. But if construction stalls, all issues could erupt in a chain reaction. Can this liability-hiding strategy save Oracle? So far, the pressure remains immense. Just two weeks ago, S&P Global downgraded Oracle’s bond rating from BBB to BBB-, just one level above junk status. If Oracle is downgraded to junk, it could trigger a massive chain reaction in its financing cycle.