A falling stock price can make a promising business look like a bargain, but the real opportunity depends on what happens next. Is the market overlooking future growth, or is it correctly identifying a problem that could take years to fix? That distinction matters for Amazon $AMZN ( ▼ 1.11% ) ), Broadcom $AVGO ( ▼ 1.88% ) , SanDisk $SNDK ( ▼ 5.97% ) , and Applied Digital $APLD ( ▼ 1.82% )

All four are connected to the expanding AI economy, yet their paths to stronger performance look very different. Amazon must prove that its enormous AI infrastructure investments can translate into stronger cash flow. Broadcom needs to turn surging demand for custom AI chips into sustainable revenue while managing customer concentration. SanDisk faces the challenge of showing that AI-driven storage demand can support earnings through the memory industry's unpredictable cycles. Applied Digital, meanwhile, must turn long-term lease commitments into operating revenue while managing substantial financing costs.

What if the stock with the biggest potential rebound isn't the one that has fallen the furthest—but the one closest to proving its business model works? 👀

We’ll examine the catalysts that could change the outlook for four AI-linked stocks, the financial warning signs that could derail their recoveries, and the milestones worth watching before deciding whether a pullback represents an opportunity or a value trap.

Why NBIS today: Nebius Group rents out AI computing capacity, so it faces the same test as Applied Digital and Amazon’s AWS below: turning heavy data-center spending into revenue that covers its costs.

Imagine setting aside $500 a month for $NBIS ( ▼ 4.98% ), buying a fixed dollar amount on a schedule, a method called dollar-cost averaging. A full five-year window is not possible here: the shares (then Yandex) were suspended from trading on Nasdaq in 2022 and only resumed as Nebius Group on Oct 21, 2024. So our example starts there. The buy happens on the first trading day on or after the 9th of each month, from November 2024 through September 2026: 23 buys, $11,500 in total.

NBIS closed at $20.00 on Oct 21, 2024, its first day back, and at $221.07 on Oct 9, 2026, a price gain of about 1,005%, or roughly 239% a year compounded over those 2 years (the steady yearly rate that would give the same total gain). In this example the 23 buys would have been worth about $44,800 at the Oct 9, 2026 close, roughly 3.9 times the $11,500 put in, and a single $10,000 invested on Oct 21, 2024 would have become about $110,500. Price only: no dividends (NBIS pays none), fees or taxes, and these are examples, not forecasts. (Figures use NBIS’s Nasdaq closes from Yahoo Finance.)

The highest close of the past 52 weeks was $286.69 on Jun 18, 2026, so the Oct 9, 2026 close sat about 23% below it. The lowest close of those 52 weeks was $73.87 on Feb 5, 2026. To keep the headline in perspective, about 1,005% is only a few points above the +1,000% line, and NBIS fell about 7.4% on Oct 8, 2026 alone, so a few days of trading could move it.

Caution: past pace rarely continues. A climb this steep over two years came with deep swings along the way, and the Oct 9, 2026 close sat well below the June 2026 high, a reminder that prices can pause even after strong runs.

Sponsored resource

In partnership with T3 Live

One type of trade I make every week…

…it’s exciting.

…it’s fast.

…it only takes you opening the trade on Thursday and closing it Friday.

It's a single 0-day options trade — the kind that can expire as soon as the next day — handed to you with the strike, the entry price, and the exit already mapped. Some weeks are quiet. Some weeks one signal like ASTS runs from 65 cents to nearly 3 dollars.

Not every trade wins — and we show you the ones that don't. But the wins are why I keep showing up.

Grab this week's free report to see how it works before the next signal drops

Get the free 0-day options report

A stock can fall even when its business is getting stronger. Sometimes, the market is worried about spending, uncertain about future demand, or waiting for proof that today's growth can translate into tomorrow's profits. Other times, investors have simply become less willing to pay a premium for a promising story.

That disconnect can create an interesting opportunity, but it can also become an expensive trap.

The key is understanding what is holding a stock back and what would need to change for investors to value the business differently. A falling share price alone does not make a company undervalued, just as strong revenue growth does not guarantee a rising stock price.

Four companies illustrate this tension particularly well: Amazon, Broadcom, SanDisk, and Applied Digital. Each has a different connection to the artificial intelligence boom, and each faces a different obstacle that could determine its next major move.

The common thread is not that all four stocks are guaranteed to rebound. It is that each has a specific business question investors need answered before a sustained recovery becomes more convincing.

Amazon's AI opportunity extends far beyond selling products online. Its cloud computing business, Amazon Web Services (AWS), provides computing power, storage, networking, and other services that businesses need to operate applications and develop AI systems.

That business is becoming increasingly important as companies invest in AI infrastructure. When customers need more computing capacity, AWS can benefit from higher usage and demand for additional services. The challenge is that Amazon must often invest in data centers, servers, networking equipment, and supporting infrastructure well before the resulting capacity generates revenue.

AWS revenue grew approximately 37% year over year to around $42 billion for the quarter. That is a strong growth rate for a business already operating at substantial scale.

The underlying opportunity is straightforward: if AWS can keep attracting customers and efficiently monetize the infrastructure it builds, today's heavy spending could support significant future revenue and earnings.

However, investors cannot evaluate Amazon solely by looking at AWS growth.

The company is expected to spend heavily on capital expenditures, with reported estimates of about $220 billion for the year. That figure should be treated as a reported estimate rather than a guaranteed final amount, and the precise definition of capital expenditure matters when comparing it with cash flow. Nevertheless, the broader issue is clear: Amazon is committing enormous resources to infrastructure before all of its expected economic benefits have materialized.

This can put pressure on free cash flow, which measures the cash remaining after operating expenses and capital expenditures. When spending rises faster than the cash generated by the business, free cash flow can decline even as revenue and operating income improve.

For a long-term investor, the important question is whether those investments produce enough additional earnings and cash flow to justify their cost.

Amazon has experienced a version of this cycle before. During the 2022–2023 cloud spending slowdown, customers became more cautious about their technology budgets, and AWS growth weakened. When demand improved, the business regained momentum. That history demonstrates how cloud spending cycles can influence the company's results, although it does not guarantee that the next recovery will follow the same pattern.

Valuation also deserves careful attention. By one estimate, Amazon trades at approximately 31 times expected earnings after excluding a substantial estimated paper gain associated with its investment in Anthropic. Such adjustments can help investors distinguish recurring operating performance from changes in the value of investments, but the exact calculation depends on the period, accounting treatment, and earnings estimate used.

What could unlock the next move? The strongest evidence would be continued AWS growth accompanied by improving returns on infrastructure spending. New data centers need to become productive assets, not simply expensive additions to the balance sheet. If revenue and operating profits begin to grow faster than the cash demands of expansion, the market may become more comfortable paying for Amazon's future earnings.

Tip: Amazon's AI story is not just about how much computing capacity AWS can sell. Watch whether that capacity produces enough incremental cash flow to justify the enormous upfront investment.

Broadcom occupies a different part of the AI infrastructure chain.

While Nvidia is widely associated with general-purpose GPUs, Broadcom supplies networking technology and custom silicon that can be designed for the specific requirements of large technology companies. These custom chips, often called application-specific integrated circuits (ASICs), can be tailored to particular workloads and system designs.

Large cloud providers and AI developers may pursue custom silicon when they want greater control over performance, power efficiency, cost, or integration with their existing infrastructure. This creates a meaningful opportunity for Broadcom as AI computing expands beyond a single chip architecture.

Reported figures point to roughly $17 billion in quarterly AI chip sales, representing growth of roughly 220% year over year, alongside about $179 billion in reported orders that customers cannot cancel. These are substantial figures, but investors should verify the reporting period and the precise definition of the order backlog before using them in a valuation model. A backlog is not the same as revenue already earned, and delivery schedules, contract terms, and customer commitments determine how quickly it becomes sales.

The investment case becomes more complicated when customer concentration enters the picture.

Broadcom's five largest customers account for more than half of its sales. That creates both an advantage and a vulnerability. Large customers can commit to enormous deployments, giving Broadcom visibility into future demand. But if one major customer delays a project, changes its chip strategy, or reduces its spending, the impact could be substantial.

Reports also describe financing arrangements connected to Anthropic, including a reported commitment of up to $42 billion in lending and support for certain equipment leases. These arrangements deserve particular scrutiny because financing a customer that then uses the funds to purchase a supplier's products can create economic interdependence.

That does not automatically make the arrangement unsound. Strategic financing can help a customer secure capacity and allow a supplier to support a long-term relationship. However, it can also concentrate risk. If the customer's business underperforms or its financing becomes difficult to sustain, the supplier may face weaker demand and potential obligations connected to its guarantees or other commitments.

For investors, the distinction between orders and independently funded customer demand is important. A large backlog is more reassuring when customers have the financial capacity to fulfill their commitments and when the supplier can deliver the products at attractive margins.

Valuation adds another layer. One estimate puts Broadcom's price-to-earnings growth (PEG) ratio of approximately 0.5. Although a low PEG ratio can suggest that a stock is inexpensive relative to expected earnings growth, the result depends heavily on the earnings forecast and the calculation method. It should not be treated as proof that a stock is undervalued, especially when future growth depends on a small group of large customers.

What could unlock the next move? Investors need evidence that Broadcom's AI demand is converting into delivered products, collected payments, and sustainable profits. Customer commitments matter, but actual execution matters more. Updates from Taiwan Semiconductor Manufacturing Company (TSMC), which manufactures chips for Broadcom, may also provide useful context about AI-related semiconductor demand, although TSMC's results cannot independently confirm the financial quality of Broadcom's own contracts.

Tip: Broadcom's opportunity is tied to the expansion of custom AI chips and networking, but the strength of its customer relationships must be weighed against the risks of concentration and financing commitments.

SanDisk offers exposure to another essential part of AI infrastructure: data storage.

AI systems do not rely exclusively on processors and high-speed working memory. They also need storage for training datasets, model files, checkpoints, application data, and the enormous volumes of information generated as AI services operate.

The distinction between memory and storage matters. Dynamic random-access memory (DRAM), including high-bandwidth memory used alongside advanced AI accelerators, provides fast temporary access to data. NAND flash, which SanDisk specializes in, provides nonvolatile storage that retains information even when power is switched off.

Both are necessary, but they serve different purposes.

SanDisk's potential opportunity comes from growing demand for NAND storage in data centers, alongside its existing markets in personal computers, smartphones, and consumer devices. Reported figures point to roughly $20 billion in sales following growth of roughly 175%, with expectations for further substantial growth in the following fiscal year. These figures need to be checked against the relevant financial statements and fiscal reporting periods, particularly because SanDisk's standalone financial history differs from that of its former parent, Western Digital.

The central question is whether AI demand can make SanDisk's business less vulnerable to the traditional boom-and-bust pattern of the memory industry.

Historically, memory manufacturers have experienced sharp swings in revenue and profitability. When demand accelerates, customers compete for supply, prices rise, and manufacturers can enjoy exceptional margins. But as production expands or customer demand weakens, prices can fall quickly. A company that looks extraordinarily profitable near the top of a cycle can see earnings deteriorate just as rapidly.

That history explains why a low price-to-earnings ratio can be misleading for a memory stock. If investors expect earnings to peak temporarily, they may assign the company a lower multiple even while reported profits are soaring.

Reports point to an increase in SanDisk's data-center spending relationships and a substantial order backlog, which could suggest that AI customers are making longer-term storage commitments. Such agreements may provide greater visibility than purely short-term sales.

However, a signed order book does not automatically eliminate cyclicality. Investors need to understand whether contracts lock in prices, specify minimum purchase volumes, allow customers to delay deliveries, or contain other provisions that affect how much revenue and profit the company can ultimately recognize.

There is another concern: a significant portion of SanDisk's recent growth came from higher prices. Higher selling prices can boost revenue and margins without a comparable increase in units shipped. If those prices reverse, earnings could fall even if demand remains healthy.

Traditional consumer markets remain relevant, too. Smartphones, computers, and memory cards still account for a meaningful part of the business. Higher storage prices can encourage customers to reduce purchases, delay upgrades, or seek cheaper alternatives.

The valuation argument therefore depends on two conditions: AI storage demand must remain strong, and the profitability generated by higher prices must prove more durable than in previous memory cycles.

What could unlock the next move? Several more quarters of stable pricing, sustained data-center demand, and healthy margins would strengthen the case that AI is changing SanDisk's earnings profile. A reversal in NAND pricing or weaker consumer demand would challenge that thesis.

Tip: SanDisk should not be judged by revenue growth alone. Track NAND pricing, gross margins, data-center demand, and the actual terms of customer commitments to determine whether today's earnings can last.

Applied Digital is the most speculative of the four companies because its investment case depends heavily on the successful delivery and financing of large data-center projects.

The company develops digital infrastructure intended to support high-performance computing and AI workloads. Its business model resembles that of a specialized landlord: it invests substantial amounts upfront to build facilities, then seeks to generate recurring revenue by leasing capacity to customers.

The attraction is the potential for long-term contracted income.

Reported figures point to roughly $36 billion in signed leases, with many agreements extending for around 15 years. It also describes quarterly revenue from operating facilities of roughly $66 million. These figures should be confirmed against company filings, and investors should distinguish total contractual commitments over the life of a lease from annual revenue that has already started to be recognized.

That distinction is critical.

A long-term lease can offer visibility into future revenue, but the full value of the agreement does not arrive on day one. Facilities must be built, equipped, commissioned, and made available to customers. Until that happens, the company may have significant financing costs without receiving the corresponding rental income.

This creates a timing problem.

Applied Digital can be committed to paying interest on billions of dollars in financing while some of its future revenue remains dependent on projects that are still being completed. If construction finishes on schedule and customers begin paying as expected, the gap between rental income and financing costs may narrow. If projects are delayed, interest expenses continue to accumulate while revenue arrives later than planned.

That is why the company's debt deserves at least as much attention as its lease backlog.

By one estimate, revenue from operating buildings falls approximately $90 million a year short of covering interest costs. This should be treated as an estimate, not as a verified figure or a complete measure of profitability. A proper assessment would also examine corporate expenses, maintenance requirements, construction commitments, debt maturities, lease commencement dates, and any additional capital required to finish projects.

A company can have positive facility-level economics and still struggle financially if interest costs and corporate obligations consume the cash generated by its operations.

Valuation is also more complicated than it is for a profitable, established business. Applied Digital's reported accounting losses make the conventional price-to-earnings ratio less useful. Data-center landlords are often evaluated using measures such as funds from operations, adjusted EBITDA, enterprise value relative to revenue or EBITDA, and the economics of their contracted leases. Each measure has limitations, particularly when a company is still building much of its planned capacity.

Some analysts compare Applied Digital with established data-center landlords such as Equinix and Digital Realty. That comparison can provide context, but the businesses are not interchangeable. Mature operators with diversified customer bases, operating assets, established financing access, and proven cash generation may deserve different valuations from a developer that still needs to complete major projects.

Applied Digital's potential upside rests on successful execution. If more campuses come online, customers begin paying rent, and recurring cash flow improves relative to interest expenses, investors may become more confident in its business model. But the same leverage that magnifies returns when projects succeed can magnify losses when financing costs rise or construction falls behind schedule.

There is also substantial short interest. That indicates that some investors are betting against the stock, but it does not establish that a rebound is imminent. A heavily shorted stock can rise sharply if expectations improve, yet it can also remain weak if the business fails to meet its obligations or milestones.

What could unlock the next move? The most important signals are on-time campus delivery, customer lease commencement, rising rental revenue, and a narrowing gap between operating cash generation and financing costs. These milestones would provide stronger evidence of progress than a rising share price alone.

Tip: Applied Digital is a high-risk infrastructure investment, not simply a discounted landlord. Before focusing on its potential valuation, determine whether its contracts, construction schedule, and financing give the company enough room to reach sustainable cash generation.

Amazon, Broadcom, SanDisk, and Applied Digital operate in different parts of the AI ecosystem, but their share prices are responding to a similar tension: the market is trying to determine whether rapid growth will translate into durable financial results.

For Amazon, the test is whether AWS growth can justify enormous infrastructure spending and eventually support stronger free cash flow.

For Broadcom, it is whether custom-chip demand and its large order book can translate into collected revenue without excessive customer concentration or financing risk.

For SanDisk, it is whether AI-related storage demand and favorable pricing can deliver earnings that are more durable than the memory industry's historical boom-and-bust cycles.

For Applied Digital, it is whether new data-center capacity can come online on schedule and generate enough recurring income to support its substantial financing obligations.

These are four different business questions, and each requires a different set of indicators. Treating all four stocks as interchangeable AI opportunities would overlook the very risks that explain why their valuations and share-price behavior can differ.

The idea of a coiled spring can be useful as a way to think about the gap between business progress and investor expectations. A stock may recover when uncertainty fades, a major milestone is achieved, or earnings exceed what the market anticipated. But there is no reliable way to know when that change will happen, and some stocks remain under pressure because the underlying business never delivers what investors expected.

That is why patience alone is not a complete strategy. Patience works best when it is paired with a sound investment thesis, a reasonable valuation, and regular checks that the business is progressing as expected.

For an investor with limited time, the goal is not to monitor every daily price movement or guess the exact bottom. It is to identify the few milestones that matter most, review them as new results arrive, and avoid committing too much capital to a thesis that has not yet been proven.

The bottom line: A pullback can create an opportunity, but it is not the opportunity itself. Amazon, Broadcom, SanDisk, and Applied Digital each have potential catalysts, yet every catalyst comes with conditions that must be met. The strongest long-term decisions come from understanding what needs to go right, what could go wrong, and whether the current stock price offers enough compensation for taking that risk.

Sponsored resource

In partnership with StockEarnings

Markets are down, but smart money is circling.

In under 5 minutes, you’ll discover three battered but fundamentally strong picks with massive upside as conditions normalize.

These aren’t flavor-of-the-month names. They're backed by long-term trends, strong leadership, and ideal entry points created by panic selling.

The dip is real. The opportunity is rare. With AI, cyclical rebounds, and broad exposure, this report gives you an actionable edge. Get in before Wall Street catches on.

Download the FREE report and position yourself for the rebound.

Want a simple next step for the $500 habit?

Our Fast Track Blueprint walks through a trusted online broker setup and a clear portfolio starting path.

Thanks for reading. This format is built to be fast to open, clear to understand, and useful enough to act on, without pretending past returns continue forever.

Caution: Past pace rarely continues. All figures here are approximate and are shown as examples, using price only (no dividends, fees, or taxes). Past performance is not a forecast; this is education, not advice.

Disclaimer: This newsletter is for informational purposes only and is not financial advice. Consult a qualified advisor before investing.