Organization & Compilation: ShenChao TechFlow

Guest: Thomas Hughes, market analyst

Host: Jessica Mitacek

Podcast Source: MarketBeat

Original Title: Wall Street Just Gave a Dire Warning. (Most Aren't Ready)

Broadcast Date: June 25, 2026

Summary of Key Points

Analyst Thomas Hughes believes that over the past 18 months, while large-scale tech companies and AI infrastructure firms have incurred about $750 billion in new debt, the related backlog of orders has accumulated to about $2.1 trillion during the same period, which is about three times the former. This suggests that the current market panic about "burning money" is more akin to upfront investments before fulfilling contracted demand. In his judgment, true revenue recognition is likely to wait until the end of 2027 to early 2028 when the first generation of AI data centers come online, hence fluctuations in the coming quarters will persist, but the long-term upward logic for large-cap tech stocks and core infrastructure companies remains intact.

Highlights of Key Insights

Market Panic and Misinterpretation of "Burning Money"

• “The market's hesitation is mainly due to worries that data center expansions and large-scale expenditures will impact these companies' debt, balance sheets, and cash flows.”
• “In the past 18 months, ultra-large-scale cloud vendors and AI infrastructure companies have added about $750 billion in debt to promote data center construction. Investors see cash flows rapidly declining from their highs, naturally causing short-term pressure, which creates headwinds for both market sentiment and stock performance.”
• “AI development carries risks, but these are execution risks. It is more likely to manifest as periodic price adjustments rather than a complete collapse in stock prices. Over an extended period, these companies will likely continue to move upward.”

Backlog of Orders as a More Critical Leading Indicator

• “What best illustrates that this is not a 'storytelling emerging tech narrative,' but rather an upgrade of the tech system, is the backlog of orders. This round of spending, debt expansion, and data center construction is not about betting on potential future business but about delivering already contracted business.”
• “This means that the long-term logic is they will convert the capital borrowed today into future revenues that are approximately three times the debt.”
• “In the coming years, we should see quite strong growth, while debt levels gradually decline, free cash flow recovers, and buybacks can accelerate again.”

Time Lag in Revenue Realization and Market Volatility

• “Overall, I think they explain it fairly well because the confirmation cycle for these contracts is inherently delayed, many of which rely on computations that will only be released in the future. You must recognize that the current existing computational capacity is already close to full, which is why prices are going up.”
• “While it looks like there is already a $2.1 trillion backlog, a significant portion of it will not be immediately recognized as revenue, and may not convert to actual income until the end of 2027 or even early 2028, after these new data centers come online.”
• “Fluctuations will continue for several quarters. The AI bubble is likely to expand further, which will create considerable upward opportunities between earnings seasons; meanwhile, the market will constantly find new reasons to worry, remain cautious, or simply take profits, leading to frequent pullbacks.”
• “In the current environment, price weakness feels more like a buying opportunity.”

Oracle's Core Position

• “For me, Oracle is one of the highest quality AI narratives.”
• “It is a large-scale cloud provider serving other large-scale cloud vendors. It provides high-capacity, high-performance computational services to other large cloud providers, Meta, various AI labs, and numerous enterprise customers. Its originally core database business is already deeply embedded in the entire cloud ecosystem, integrated into the systems and networks of major cloud providers.”
• “From the current trends, I believe it has already been establishing a quite clear bottom. The market is seriously oversold, and the response to risk has been somewhat excessive. The current structure is actually quite conducive to a rebound; it just needs a real catalyst to turn sentiment around.”

Chips and the AI Supercycle

• “Under the current circumstances, Micron's production capacity is already scheduled until the end of next year, and this earnings report may even push that timeline to 2028. This will continue to support price expectations and directly drive its current business performance. What supports it is not only sales but also the prices it can achieve.”
• “The chip industry itself is in a massive supercycle. Initially driven by inventory normalization, it has been further enhanced by AI. As long as the supply side cannot increase capacity to meet demand, this cycle will continue for many years.”
• “Investments in AI will create new capacity and new technologies, and these new outputs will, in turn, improve efficiency, enabling the industry to invest more funds for the next round of upgrades. In other words, every round of investment will drive technological progress, and technological progress will again drive the next round of larger investments, perpetually. At least for now, this flywheel effect of AI likely does not have a clear endpoint.”

Long-Term Judgment

• “This cycle is far from over. I believe that technology stocks, especially leading blue-chip operators, have a long way to go with good days ahead. AI will drive technological changes, but these companies are already in the most advantageous positions to continue moving forward with these changes.”
• “We haven't even fully completed the first generation of AI data centers. At the current pace, the first generation of data centers will only start to come online gradually next year.”
• “Once we truly enter the application phase and companies start to monetize these technologies, we will see revenue and profit begin to emerge.”
• “They have sufficient funds, scale, and execution capabilities to truly make this happen.”

Opening: Why the Market Suddenly Became Tense

Host Jessica Mitacek: In just two days, large tech stocks have evaporated over $2.5 trillion in market value. Market headlines are almost all saying that these tech giants' spending on AI has become too large to ignore, but is this round of sell-off truly an overreaction? From a long-term perspective, is there a deeper and more important story behind this? Today in the program is MarketBeat analyst Thomas Hughes, who will break down what is currently happening with ultra-large-scale cloud vendors, and will also discuss three companies worth paying attention to in the future. Thomas, let's start with how the market is reacting right now.

Thomas Hughes:

A few weeks ago, the market just peaked, and it seems to be struggling to regain upward momentum. The market's hesitation is mainly due to fears that data center expansions and large-scale expenditures will impact these companies' debt, balance sheets, and cash flows.

Host Jessica Mitacek: This pullback is indeed very evident on the heatmap for the past 7 days and is almost entirely concentrated in the large tech sector. For instance, Oracle has dropped over 14% in the last 7 trading days, Google is down nearly 4%, and the entire big tech area is almost all in red. The problem is that these companies had just gone through a sharp rally before this, so what everyone wants to know now is what triggered this pullback. A common saying is "burning money," especially in the second quarter earnings reports, ultra-large-scale cloud vendors disclosed significant spending numbers. What is your take on this issue?

Thomas Hughes:

If we look roughly, in the past 18 months, ultra-large-scale cloud vendors and AI infrastructure companies have increased about $750 billion in debt to promote data center construction, which includes some financing actions that bring dilution effects. On one hand, they are using free cash flow, while on the other hand, the new debt is eroding free cash flow because cash must be used to maintain the debt.

Previously, these companies have been engaging in large-scale stock buybacks for a long time, but the money that was originally intended for buybacks is now being diverted to handle debt, which is why cash flow has been significantly constrained. Investors see cash flows rapidly sliding from their highs, and in the short term, it naturally causes anxiety, creating headwinds for both market sentiment and stock performance.

Host Jessica Mitacek: I saw a chart on Yahoo Finance today that just compares this change. The issue is not just how much was spent in the last quarter, but that these companies have been investing large amounts of capital in AI expansion consistently for several quarters. The real big question is, will these investments yield returns in the future? Is this an overly aggressive, high-risk bet, or will it ultimately generate significant returns? Could such cash consumption put ultra-large-scale cloud vendors at substantial risk?

Thomas Hughes:

There is risk because the amounts are indeed substantial. But ultimately, this is more like execution risk. They must build these systems, launch them, and then successfully monetize them.

However, compared to early-stage emerging tech companies, this type of risk is not that high. Because this is not an unproven demand that still needs to build a business model from scratch, but rather an evolution of the existing technology system. Almost all leading companies are already involved in this transformation and are collaboratively driving it forward. From this perspective, the risk is significantly mitigated.

The technical direction is already clear, the problem is not whether to prove that this path works, but whether they can build it well, get it running, and ultimately make money. Therefore, my view is: there are risks, but these are execution-level risks. It is more likely to manifest as periodic price adjustments rather than a complete collapse in stock prices. Over the long term, these companies are likely to continue moving upward.

Host Jessica Mitacek: The market has become accustomed to these companies continually leading the pack, and accustomed to their continuous buybacks, expansions, and consistently outperforming growth. Now they are entering a new investment cycle, needing to heavily invest in AI infrastructure first, which indeed makes investors uneasy. Moreover, I want to highlight that many retail investors feel stocks like Google and Nvidia are quite "boring," but the reality is that almost all investors indirectly hold these assets. Even if you don’t buy stocks yourself, if you have a 401(k) or various ETFs, you usually can't avoid them, so everyone wants to know what this means for their portfolios.

Thomas Hughes:

Exactly, almost everyone holds these stocks. They are one of the biggest sources of income and cash flow, and their growth paths are very clear. Even if investors do not hold directly, they are likely to hold them through ETFs or funds. So many people think they have no exposure, but that's not necessarily the case.

Beyond "Burning Money," The More Important Issue is Order Backlog

Host Jessica Mitacek: Do you think the current "burning money" is a risk that investors should truly be worried about, or does it actually point to a deeper, more long-term logic?

Thomas Hughes:

I believe it reflects a more long-term logic. The most telling sign that this is not a "storytelling emerging tech narrative," but rather an upgrade of the tech system, is the backlog of orders.

This round of spending, debt expansion, and data center construction is not about betting on potential future business but rather about fulfilling already contracted business. While debt is expanding, these companies' backlogs are also growing at a faster pace. Currently, from ultra-large-scale cloud vendors to AI infrastructure firms, from chips to foundational hardware for data centers, and computational resource leasing, the total backlog along the entire supply chain is approximately $2.1 trillion, nearly three times the scale of new debt.

This means that the long-term logic is they will convert the capital borrowed today into future revenues that are roughly three times the debt. In the coming years, we should expect quite strong growth, while debt levels gradually decline, free cash flow recovers, and buybacks can accelerate again.

Moreover, it's important to note that these backlogged orders themselves are signed for several years. We believe this $2.1 trillion will likely be enough to support the next 3 to 5 years. Once these contracts are fulfilled, new contracts will come in. By then, the data centers will be built, and there won’t be a need to incur new debts, leaving just rolling new contracts with better profit margins. Thus, from an AI long-term perspective, I think the overall outlook is very strong.

Host Jessica Mitacek: There is a crucial point here. The backlog of orders you mentioned is not just those companies that actually build data centers or sell equipment and components, right? We know that companies like Google, Amazon, and Meta are investing significant funds, effectively transferring money to those "shovel sellers" in the infrastructure space. So, does this backlog of orders exist not only in the infrastructure supply chain but also within the ultra-large-scale cloud vendors themselves?

Thomas Hughes:

Of course, it does, and it spans the entire industry chain. It is not only reflected in the GPUs that support AI but also in the components and data center infrastructure that assemble GPUs, as well as in the actual computational capacity itself.

These large ultra-large-scale cloud vendors and many large software service companies have long been signing contracts and are continuing to sign contracts for future computational capacity. The real driving force behind the entire growth story is their demand for computational capacity; it is this demand that, in turn, drives infrastructure development.

So, the stage we are currently in is essentially an infrastructure expansion phase. The true long-term investment line will shift from "building capacity" to "utilizing capacity." Once capacity starts to be widely used, the backlog of orders will significantly convert to revenues and profits.

When Will Orders Convert to Revenue

Host Jessica Mitacek: There is no doubt that the market's demand for AI computational space is real. Whether it is enterprises, companies, or ordinary investors, AI is being used more frequently. Since the demand is indeed growing, it’s easy to understand why these ultra-large-scale cloud vendors are expanding massively. Google is a very typical example; it is investing heavily in AI, but what investors are most concerned about is: how will this money translate into Google’s own revenue in the future? What will its monetization path look like?

Thomas Hughes:

Google is a great case study. You can see from the chart that it is not just Google; the entire Mag 7, including Nvidia, has experienced a decent pullback. But to me, this feels more like a very normal market mechanism during the earnings seasons. After the last earnings round, the market rose too quickly and sharply, creating conditions for a correction.

So, I think the market is more like recharging for the next rush. In the next two to four weeks, there will be a new earnings window, and these companies' reports will likely still be good and at least continue to validate a few trends: first, capital expenditure remains robust; second, the construction of data center capacity is still advancing; and third, there is still demand from the end market for computational capacity.

Host Jessica Mitacek: I've heard a common saying lately: these earnings reports will disclose AI expansion spending, thus benefiting the “AI investment story,” but the issue is whether these ultra-large-scale cloud vendors have clearly explained during their earnings calls how today’s investments will truly translate into their own revenue in the future?

Thomas Hughes:

I believe that overall, they actually explain it fairly well, because the confirmation cycles of these contracts are inherently delayed, many of which rely on future computational capacity to be released. You have to recognize that the current existing computational capacity is already close to full, which is also why prices are rising.

Thus, many contracts are actually tied to future computational capacity and future technology. A significant portion of data centers will be built based on next-generation products, including newly launched platforms and products like AMD's MI450 series that have yet to be fully implemented. In other words, while it may seem that there is already a $2.1 trillion backlog, a large portion of it will not immediately be recognized as revenue; it may take until the end of 2027 or even early 2028 before these new data centers come online and start to significantly convert to actual revenue.

Host Jessica Mitacek: This timing is very important for investors. Because if the realization of revenue still requires a wait of one to two years, will the market continue to oscillate during that waiting period? Many large tech stocks have likely evaporated in value precisely because investors are reluctant to bear the risks during this wait. Do you think concerns surrounding high expenditures and slow realization of revenue will persist over the next one to two years?

Thomas Hughes:

I believe fluctuations will continue for several quarters. The AI bubble is likely to further expand, which will continue to create considerable upward opportunities between earnings seasons; at the same time, the market will constantly find new reasons to worry, remain cautious, or simply take profits, leading to frequent pullbacks.

But in the current environment, I believe price weakness feels more like a buying opportunity. Because the catalysts for driving stock prices up in the future are not only today's performance and guidance but also the process of these backlog orders eventually converting to cash flows and profits. Once this step begins to be realized, stock prices will be supported over a longer time frame.

Stock One: Oracle

Host Jessica Mitacek: We just talked about Google. Next, let's discuss a few companies you are particularly watching, one of which is Oracle. You recently wrote an article about its recent pullback, and it is a representative of high-investment expansion companies.

Thomas Hughes:

Oracle can be considered one of the most typical representatives in this story. It is indeed very actively using debt leverage, but at the same time, its backlog of orders is also among the top in the companies I observe.

To me, Oracle is one of the highest quality AI narratives. Because it is not only upgrading from a traditional tech company to a modern cloud and AI company, but also transforming from a relatively niche player to a key blue-chip company at the core of the AI and data center ecosystem. You could say Oracle is a large-scale cloud provider serving other large-scale cloud providers.

It provides high-capacity, high-performance computational services to other large cloud providers, Meta, various AI labs, and numerous enterprise customers. Moreover, this is only a part of its cloud business; its originally core database business is already deeply embedded in the entire cloud ecosystem, integrated into the systems and networks of major cloud providers. It is also one of the most widely used and easily accessible databases globally, making it a very crucial foundational role in the entire AI infrastructure and industry chain.

Host Jessica Mitacek: Compared to Google or other Mag 7 companies, does Oracle carry higher risks? After all, its starting balance sheet isn't as robust as those super giants.

Thomas Hughes:

I don't think its risk is higher. In fact, before this round of expansion began, its balance sheet was already quite healthy, and it had been actively repurchasing stocks. Some financial indicators may raise concerns, but they have been offset by its strong cash flow and aggressive buybacks.

I think the biggest worry now is that debt is expanding, which may create a drag over the next 12 to 18 months, possibly even a bit longer. But as I have mentioned before, starting next year, it will gradually confirm backlog orders as revenue and continue to amplify that over the subsequent quarters, which will quickly digest the debt. So, I think the force pressing down on the stock price this year will be corrected in the coming years, ultimately reflected in the stock price moving upward.

Host Jessica Mitacek: If we just look at the chart, Oracle has fallen about 15% in recent days, and this volatility has been ongoing for the past year. It hit a high last September and has been in a high volatility range ever since. So what investors are most concerned about now is: is this the beginning of a deeper downturn, or is it already at a bottom worth entering?

Thomas Hughes:

In my opinion, this is not the beginning of a downtrend; it is a very clear buying opportunity. That large surge in Oracle stock last year essentially resulted from a rapid swelling of backlogged orders. The subsequent decline in stock price was mainly because the market began to worry about the debt issue.

However, from the current trend, I believe it is already building a quite clear bottom. Although it has been correcting over the past few weeks, I see a typical head and shoulders bottom pattern. The market is severely oversold, and the response to risk has been somewhat excessive. The current structure is actually very suitable for a rebound; it just needs a real catalyst to turn sentiment around.

Oracle belongs to the group of companies that disclose their earnings in the mid-cycle of the earnings period; it recently released an earnings report, so there are still a few weeks until the next earnings report, and it will be after other ultra-large-scale cloud vendors. But in my opinion, other giants' earnings reports will inherently become catalysts for Oracle because part of the expenditures disclosed by them is meant to flow to Oracle to support Oracle's own large-scale expansion.

Stock Two: Micron and the Chip Chain

Host Jessica Mitacek: Although Oracle will not disclose new earnings soon, the earnings report that the market is currently most focused on is actually Micron. As we record this program, it hasn't closed yet, so the report hasn't come out. However, everyone is clearly looking forward to seeing how much of this massive spending is flowing into storage and chip companies like Micron. So, is there any connection between this earnings report and Oracle or the broader AI investment story?

Thomas Hughes:

Currently, a significant amount of funds is flowing into the chip sector, and it's not limited to just GPUs or companies like Micron focusing on memory chips. Micron's stock price recently hit a new high this week, retracing before the earnings report, but to me, the chart still appears very strong.

The market is currently expecting it to continue showcasing extremely high-demand levels, and it may even delay its schedule for meeting capacity demands further. Under the current circumstances, Micron's production capacity is already scheduled through the end of next year, and this earnings report may even push this timeline to 2028. This will continue to support price expectations and directly drive its current business performance. What supports it is not just sales but the prices it can achieve.

If we broaden the view to the entire chip industry, AI is driving demand across the board. Almost all chips you can think of will be utilized in data center construction. Looking further ahead, AI applications will extend to the Internet of Things and various AI application scenarios, because GPUs are just the brain responsible for computation; in addition, many chips are needed to connect GPUs, link servers, connect data centers, and support power control devices, actuators, as well as all kinds of servers that enable the entire IoT to operate.

Ultimately, AI will lead to "AI in the physical world" and the Internet of Things, utilizing this computational brain to allow machines to remotely execute work globally.

Stock Three: How Long Can the AI Supercycle Last

Host Jessica Mitacek: This also explains why more and more investors are using AI, as once they start using it, they find it does indeed create value. However, the next major concern for the market is: how long can this super growth cycle continue? Especially in the chip sector, will this high prosperity reach a peak soon?

Thomas Hughes:

The chip industry itself is currently in a massive supercycle. It was initially driven by inventory normalization and has been further reinforced by AI. As long as the supply side cannot increase capacity enough to meet demand, this cycle will continue for many years.

If we look specifically at memory chips, it will likely take about another year for the industry to clearly ramp up production capacity to meet current demand. This means that in the next year or so, chip manufacturers will likely continue to enjoy good times.

But looking further ahead, AI is forming a flywheel effect. Investments in AI will create new capacity and new technology, and these new outputs will, in turn, improve efficiency, allowing the industry to invest more funds for the next round of upgrades. In other words, each round of investment will drive technological progress, and technological progress will again drive the next round of larger investments, perpetually.

At least for now, this flywheel effect of AI likely does not have a clear endpoint. The entire industry is still in a very early stage, we haven't even truly completed the first generation of AI data centers. At the current pace, the first generation of data centers is set to gradually come online next year.

Host Jessica Mitacek: This also ties back to another thing that the market is currently most concerned about. Everyone knows that money is being invested in the first generation of AI data centers, but the question is, by the time these facilities are built, will they already be outdated? Will they be fully utilized as ultra-large-scale cloud vendors expect, or will the market have turned to a newer generation of technology before they are completed? The potential for these infrastructures to quickly depreciate is also a significant concern for investors right now. How would you respond to this worry?

Thomas Hughes:

In the tech industry, once data centers are built, there will certainly be better products that come along later, but the arrival of new generations of technology takes time. You have to see that this round of AI data center enthusiasm has actually been brewing for several years, yet we still have not seen large-scale results materialize.

We have certainly seen some initial results, such as extremely strong demand for infrastructure products from some companies, but those data centers themselves are not truly completed yet. Therefore, large-scale application of technology is still to come.

Once we truly enter the application phase and enterprises begin to monetize these technologies, we will start to see revenue and profits emerge; during this process, funding will gradually shift to the next generation of products. By that time, a new supercycle may form again.

Host Jessica Mitacek: So your conclusion is still positive about the Mag 7, about ultra-large-scale cloud vendors, and also about the long-term outlook of the entire AI industry chain. You don't believe this story is nearing its end, nor do you think the risks have become so significant that they could disrupt the future of these companies.

Thomas Hughes:

This cycle is far from over. I believe that technology stocks, especially leading blue-chip operators, have a long way to go with good days ahead. AI will drive ongoing technological changes, but these companies are already in the most favorable positions to continue moving forward with these changes. They have sufficient funds, scale, and execution capabilities to truly make this happen.