AAOI Stock: The 800G Capacity Ramp Against the Dilution Bill

Applied Optoelectronics (NASDAQ: AAOI) is one of the very few American companies that can make an optical transceiver and the laser chip inside it under the same roof. In an AI buildout where the biggest bottleneck has quietly moved from GPUs to the optics that connect them, that is a genuinely valuable position.

It is also an expensive one. AOI has spent the first half of 2026 raising equity at a pace that would be alarming in almost any other context, and pouring it into factory floor space and production equipment. The company has record revenue, a $200 million first order for its newest product, and demand it openly says it cannot meet. It also has a widening GAAP loss, negative operating cash flow, and roughly 12.5% more shares outstanding than it had at the start of the year.

This article takes that tension seriously rather than resolving it in advance. The question is not whether AAOI stock goes up. It is whether the capacity ramp reaches profitable scale before shareholders are diluted out of the upside.

1. Key Metrics at a Glance

Metric	Value
Share price	$105.17
Market capitalisation	$8.93 billion
52-week range	$18.50 to $233.67
Q2 2026 revenue	$191.9 million (+86.3% YoY)
Q2 2026 GAAP net loss	$(22.8) million, or $(0.28) per share
Q2 2026 non-GAAP net income	$5.5 million, or $0.06 per diluted share
Q3 2026 revenue guidance	$255 million to $290 million
FY2026 revenue expectation	Around $1.1 billion
FY2025 revenue	$455.7 million
Cash and equivalents (30 Jun 2026)	$508.8 million
Convertible senior notes	$129.1 million carrying value
Shares outstanding (30 Jun 2026)	84.4 million, up from 75.0 million
Analyst average price target	$140.68 (13 analysts)

Data as of 20 September 2026. Verify before publishing.

2. What Applied Optoelectronics Actually Sells

AOI makes optical transceivers, the modules that convert electrical signals into light and back again so that data can travel across fibre inside and between data centres. It sells into three end markets, and the mix has shifted dramatically over the past two years.

In the second quarter of 2026, datacenter products generated $107.7 million, or 56.1% of revenue. The CATV segment, which supplies cable and broadband operators, delivered $80.6 million, or 42.0%. Telecom contributed a rounding error at $3.4 million.

What separates AOI from a contract assembler is vertical integration. The company fabricates its own laser chips rather than buying all of them from the merchant market. On the second-quarter call, management described in-house laser manufacturing as positioning the company to support both near-term customer needs and longer-term growth, and it is expanding that laser capacity in Texas specifically.

That matters more than it sounds. Management was blunt on the same call that the laser “is the biggest bottleneck right now for the transceiver business,” and that lasers, DSPs and TIAs all remain constrained, especially for 1.6 terabit products. A transceiver maker that owns part of its own laser supply is not immune to that shortage, but it is less exposed to it than one that owns none.

The CATV business is not a legacy afterthought

It is tempting to read AOI purely as an AI story and ignore the cable segment. The numbers argue otherwise. CATV revenue of $80.6 million in Q2 was up 43.8% year over year and 20.6% sequentially, driven largely by 1.8 GHz amplifier shipments to the company’s largest cable customer. Management guided CATV to between $100 million and $110 million in Q3 and continues to target over $325 million annually.

For an investor, this segment is doing something useful: it is funding a portion of the fixed cost base while the datacenter ramp is still loss-making at the GAAP line. It is also a genuinely different demand cycle, which is worth something in a portfolio context.

3. Why Optics Became the AI Bottleneck

Training and inference clusters have grown large enough that the connections between accelerators now constrain the system as much as the accelerators themselves. As cluster size grows, more of the traffic has to travel distances that short-reach copper cannot serve reliably at the required bandwidth, which pushes the interconnect toward optics and toward higher per-lane speeds.

That is the mechanical reason 800G and 1.6T transceivers are ramping so violently. It is not a marketing cycle. It is a physical constraint on how large a coherent compute cluster can be built.

The consequence for suppliers is unusual: demand is not the variable being managed. Capacity is. AOI said explicitly on its Q2 call that “our near-term revenue is bounded almost entirely by production capacity and key component availability,” and that forecast demand continues to outpace production capacity through mid-2027.

This is the single most important framing in the whole investment case. When revenue is capacity-bound rather than demand-bound, capital spending stops being optional and becomes the business.

4. The Q2 2026 Result in Detail

AOI reported its second quarter on 6 August 2026. Revenue was $191.9 million, up 86.3% year over year and 27.1% sequentially, and the fifth consecutive quarterly record.

The bottom line is where it gets interesting, and where a casual reading goes wrong.

GAAP: operating loss of $(24.7) million and a net loss of $(22.8) million, or $(0.28) per basic share. GAAP gross margin was 27.7%, down from 30.3% a year earlier and from 29.1% in the first quarter.

Non-GAAP: net income of $5.5 million, or $0.06 per diluted share, against a consensus estimate of roughly $0.01. Non-GAAP gross margin was 29.8%, up sequentially from 29.2%.

So the company was profitable on a non-GAAP basis for the first time in this cycle, and considerably less profitable on a GAAP basis than a year earlier. The gap between the two is roughly $28 million in a single quarter, and it is not noise. It is share-based compensation, the accounting treatment of the Amazon warrant, and the other items that get added back. An investor who tracks only the adjusted figure is tracking a number that excludes a real transfer of value away from existing shareholders.

The honest reading of Q2 is that operations are improving and the reported result is getting worse. Both are true simultaneously, and that is exactly what a heavily funded capacity ramp looks like on the way up.

Product mix is turning over fast

The datacenter revenue split in Q2 tells the story of a product transition happening in real time:

• 100G: 38.3% of datacenter revenue
• 200G and 400G: 45%
• 800G: 11.9%

800G revenue more than doubled sequentially in the quarter, and management expects it to grow roughly fivefold in Q3. The older generations are on their way to becoming the minority of the business within a year.

The guidance was a beat and a miss at the same time

Q3 2026 guidance is revenue of $255 million to $290 million, a midpoint of $272.5 million and roughly 130% year-over-year growth at that midpoint. Against a consensus figure of about $278 million, the midpoint landed slightly below.

Non-GAAP EPS guidance of $0.11 to $0.26 sits against a consensus near $0.28, so even the top of the range only just reaches what analysts had modelled. Non-GAAP gross margin is guided to 29% to 30.5%, and the diluted share count assumption is roughly 92.8 million.

That last number deserves a pause. The company guided Q2 on a share count near 80.7 million. One quarter later the planning assumption is 92.8 million. The dilution is not a background detail in this story; it is embedded in the guidance itself.

There is also a specific, disclosed headwind in Q3. Management expects a decline in 100G business because one customer cannot source enough 100G switches to meet its own initial forecast, trimming an estimated $20 million to $25 million from sales. That is a supply problem at the customer, not evidence that demand for AOI’s product disappeared, and the company said so directly.

5. The Capacity Ramp: What Is Actually Being Built

This is the part of the story that is concrete and checkable, and it is genuinely impressive.

• Combined 800G and 1.6T capacity is approaching 200,000 units per month, up from nearly 100,000 per month at the end of Q1 2026.
• The target is over 650,000 pieces per month by the end of 2026.
• The 2027 target is over 930,000 pieces per month.

Physically, that expansion is anchored in Sugar Land, just outside Houston. Through a combination of purchases and leases, AOI has expanded its Texas footprint to over 1.6 million square feet in the greater Houston area. A new 210,000 square foot facility was expected to begin initial production late in the third quarter, dedicated entirely to 800G and 1.6T transceivers.

On 1.6T specifically, management expected full qualification of its first 1.6T product with a customer within weeks of the August call, with shipments starting later in the third quarter. The company received its first volume order for 1.6T transceivers, announced at $200 million, with deliveries very late in Q3 ramping into Q4. Management expects more than $70 million of 1.6T revenue in the fourth quarter.

A target going from 500,000 units per month at the end of 2025 to over 650,000 in the most recent update is a raise, not a miss. It is worth saying that plainly, because the bear framing on AOI often assumes the targets are always slipping. On this specific metric, over this specific period, they have moved up.

6. The Funding Question

Here is the other half of the ledger, and it is large.

Operating cash flow. Net cash used in operating activities was $(73.8) million for the six months ended 30 June 2026. That is an improvement on the $(116.4) million used in the same period of 2025, but it is still cash going out the door from operations while revenue is at a record.

Capital spending. In the same six months, AOI deployed $(335.1) million on property, plant and equipment and a further $(289.7) million in advance equipment payments. That is roughly $625 million of capital commitment in half a year against $343.1 million of revenue in the same half.

Where the money came from. Equity. During the first half, public stock offerings generated net proceeds of $1.028 billion. Cash rose from $216.0 million at the end of 2025 to $508.8 million at 30 June 2026 despite that enormous capital outlay.

The ATM. On 14 May 2026, AOI entered an equity distribution agreement with Raymond James and Needham covering up to $600 million of common stock, with a 2% commission to the agents. By the time of the August earnings call, management confirmed the company had raised $538.8 million net of commissions and fees under that programme. In other words, roughly 90% of a $600 million facility was consumed inside three months of it being opened.

The dilution. Shares outstanding went from 75.0 million at 31 December 2025 to 84.4 million at 30 June 2026. That is an increase of about 12.5% in six months, and the Q3 guidance share count of 92.8 million implies the trend did not stop at the half.

Set against that, the balance sheet is not fragile. Cash of $508.8 million, total other debt of roughly $58.9 million, and convertible senior notes with a carrying value of $129.1 million is a comfortable position on paper. The converts carry a fair value of $445.1 million because the stock has appreciated so much, which is its own form of future dilution rather than a cash liability.

The question is not solvency. It is arithmetic. AOI is issuing equity to buy equipment, and the equipment only justifies the issuance if the resulting capacity converts into margin fast enough to outrun the growing share count. Revenue per share, not revenue, is the metric that matters here.

7. Customer Concentration and the Amazon Warrant

AOI’s customer base is extraordinarily narrow, and this is disclosed rather than inferred.

In the fiscal 2025 annual report, Digicomm accounted for 53.1% of revenue and Microsoft for 28.8%, a combined 81.9%. The top ten customers represented 96.6% of total revenue. In the second quarter of 2026, the top three customers came in at 42%, 26% and 24% of quarterly revenue, with the top ten at 99%.

There is no diplomatic way to describe that. Two customer decisions could halve this business.

The Amazon relationship is the other piece investors frequently misread. On 13 March 2025, AOI issued a warrant to an Amazon subsidiary to purchase up to 7,945,399 shares at an exercise price of $23.6956, with a ten-year term. At issuance, 1,324,233 shares were exercisable. The remaining 6,621,166 shares vest over the term and are contingent on aggregate purchases by Amazon and its affiliates reaching $4 billion of AOI product.

That distinction is the whole point. The warrant is a milestone-linked incentive, not a purchase commitment. Amazon has not promised to buy $4 billion of anything. It has been given a reason to, and a discounted claim on AOI equity if it does. The grant-date fair value was determined at $12.64 per share, and the company recognises that value as a reduction of revenue from Amazon as goods are transferred, which is part of why GAAP and non-GAAP results diverge.

Read correctly, the warrant is simultaneously the strongest signal in the bull case and a direct cost to existing shareholders.

8. Competition and the Supply Chain

The listed comparables in optical transceivers and lasers are Coherent (COHR) and Lumentum (LITE), both of which compete directly in the optics stack and both of which have rallied alongside AAOI through 2026 as the optics trade has broadened. AOI is the smaller and less diversified of the three.

The larger competitive fact sits offshore. Zhongji Innolight holds roughly 27% of the datacenter optical transceiver market, which makes it the reference point for pricing in the category.

That brings us to the most misunderstood item in the AAOI bull case.

The China transceiver restriction is a draft, not a rule

Reuters reported on 4 August 2026 that the FCC is drafting a rule that would ban US imports of new Chinese optical transceiver models used inside AI data centres, with the stated concern being the potential for data theft or disruption. Officials reportedly want it to take effect before the end of the year, and the measure would benefit US manufacturers including Coherent and Lumentum.

Every one of those verbs matters. The rule is being drafted. It has not been proposed formally, it has not been approved, and reporting on it explicitly cautioned that the agency could still modify or shelve it. Existing installed equipment would likely be unaffected even if it proceeds.

An investor who is underwriting AAOI on the assumption that Chinese competition is about to be legislated out of the US market is underwriting a news report about an unpublished draft. It is a real and potentially large optionality. It is not a fact in the model.

Indium phosphide: a constraint that is currently not binding

Lasers for high-speed optics are built on indium phosphide substrates, and the supply of those substrates has been a recurring worry across the sector. On this, AOI’s answer was unusually specific: the company said it has secured supply into the following year, is not currently limited by substrate capacity, and sources from two suppliers in Europe, two in Japan and three in China.

Worth noting the geography. A supply chain that runs partly through China is a slightly awkward foundation for a thesis built on restricting Chinese components, and that tension is a legitimate thing to hold in mind.

9. Co-Packaged Optics: The Next Leg

Co-packaged optics, or CPO, moves the optical engine onto the same package as the switch silicon instead of leaving it in a pluggable module at the faceplate. It reduces power per bit and shortens the electrical path, and it changes what a company like AOI sells: fewer complete transceivers, more lasers and light sources feeding somebody else’s package.

AOI’s position here is early but real. Several customers have already qualified its lasers and ELSFP modules. Production is very limited today, with a ramp anticipated later in 2026 and into 2027, building toward roughly 400,000 pieces per month in 2028.

The margin case is the interesting part. Management targets gross margin of about 55% to 65% on CPO lasers and more than 50% on the modules, against a long-term goal of returning datacenter transceiver non-GAAP gross margins to around 40%.

If those numbers prove durable, CPO is not a side project. It is the answer to the question this entire article is built around, because a higher-margin product mix is the only mechanism by which the capacity investment outruns the share count.

10. Valuation: What the Market Is Paying

At $105.17 and a market capitalisation of $8.93 billion, against a company expectation of roughly $1.1 billion of revenue in 2026, AAOI trades at approximately 8.1 times current-year sales.

For context on what that price embeds, the stock has traded between $18.50 and $233.67 over the past 52 weeks. It is currently well below its high and enormously above its low. The average analyst price target across 13 analysts is $140.68, with the most bullish targets materially higher.

Eight times sales is not an absurd multiple for a company growing revenue at 86% year over year with capacity sold out through mid-2027. It is also not a multiple that leaves room for the ramp to slip. Three specific things are already priced in at this level: that capacity reaches the 650,000 unit target, that 1.6T converts from a $200 million order into recurring revenue, and that gross margin expands toward the 40% goal rather than staying near 30%.

Note also that the multiple is calculated on a share count that is actively growing. If the company issues another 10% of its shares to fund the 2027 capacity target, the sales-per-share denominator moves against the holder even if revenue lands exactly where guided.

Is the valuation justified? On the base case, roughly. It is a fair price for successful execution and a poor price for anything less. There is no margin of safety embedded here, and investors should be clear-eyed that they are paying for the ramp working rather than being compensated for the risk that it does not.

11. Bull and Bear Case

🐂 Bull Case

• Revenue is capacity-bound, not demand-bound, with forecast demand exceeding production through mid-2027. That is the rarest and most valuable position a component supplier can occupy.
• Capacity is on a verified upward path: approaching 200,000 units per month now, targeting over 650,000 by end-2026 and over 930,000 by end-2027.
• 1.6T has moved from concept to a $200 million first volume order, with more than $70 million of expected revenue in Q4 2026.
• Vertical integration into laser fabrication in Texas is a structural advantage precisely when lasers are the industry’s stated bottleneck.
• CPO offers a path to 55% to 65% gross margin on lasers, which is the mechanism by which scale finally converts into profit.
• The CATV segment is growing 43.8% year over year and guided to over $325 million annually, providing a second, uncorrelated revenue stream.

🐻 Bear Case

• GAAP losses widened, not narrowed, during a record quarter. The operating loss of $(24.7) million came on the highest revenue in company history.
• Shares outstanding rose 12.5% in six months, a $600 million ATM was roughly 90% consumed within three months, and the Q3 guidance share count is already 92.8 million.
• Operating cash flow remains negative at $(73.8) million for the half, so every dollar of the $625 million capital programme was externally funded.
• Customer concentration is extreme: two customers were 81.9% of fiscal 2025 revenue, and the top ten were 99% of Q2 2026 revenue.
• The Amazon warrant is milestone-linked, not a purchase commitment, and it dilutes shareholders while reducing recognised revenue.
• The China import restriction is an unpublished FCC draft that could be modified or shelved, yet it carries real weight in how the stock is being priced.
• Q3 guidance came in below consensus on both revenue midpoint and non-GAAP EPS, which is not what a stock at eight times sales is priced for.

Base case. AOI hits the capacity targets approximately on schedule, 1.6T revenue arrives roughly as guided, gross margin grinds toward the mid-30s rather than jumping to 40%, and the company issues more equity to fund the 2027 expansion. Revenue compounds impressively and per-share value compounds far less impressively. That is a good business and a merely acceptable stock at the current price.

12. What to Watch Next

• Q3 2026 results, expected in early November 2026 though the date is not yet formally confirmed. The single most important line is gross margin, not revenue.
• 1.6T qualification and first shipments. Confirmation that the $200 million order is actually shipping, and whether Q4 1.6T revenue reaches the stated $70 million plus.
• Capacity verification. Whether monthly output is credibly tracking toward the 650,000 unit end-of-year target.
• Further equity issuance. Any new ATM facility, or heavy drawdown of an existing one, tells you the funding gap is still open.
• The FCC rule. Whether the Chinese transceiver restriction is formally proposed, and on what terms.
• Customer concentration disclosure in the next annual filing, and whether the Amazon relationship has begun to show up as meaningful revenue.

13. Final Thoughts

Applied Optoelectronics matters because it is building domestic capacity in the one part of the AI supply chain that is genuinely short, and because it owns a piece of the laser supply that everyone else is fighting over. That is not a narrative. It is a factory in Sugar Land with 1.6 million square feet and a $200 million order in hand.

The biggest opportunity is the margin structure on the other side of the ramp, particularly in co-packaged optics, where the targeted gross margins are roughly double what the company earns today.

The biggest risk is not that the company fails. It is that it succeeds on the income statement and disappoints on a per-share basis, because the equity issued to build the capacity grew faster than the profit the capacity produced. That is a quieter way to lose money than a blow-up, and it is the specific thing to monitor.

For a long-term investor, the sensible posture is to treat AAOI as a position sized for volatility, judged quarter by quarter on gross margin and share count rather than on revenue headlines. The revenue is going to keep setting records. That was never the question.

FAQ

Is AAOI profitable?

Not on a GAAP basis. In the second quarter of 2026 the company reported a GAAP net loss of $(22.8) million, or $(0.28) per share, on record revenue. It did report non-GAAP net income of $5.5 million, or $0.06 per diluted share, its first non-GAAP profit of this cycle. The roughly $28 million gap between the two figures is real economic cost, mostly share-based compensation and the accounting treatment of the Amazon warrant.

Why is AAOI issuing so much stock?

To fund a capacity expansion that its operating cash flow cannot cover. In the first half of 2026 the company spent about $625 million on equipment and advance equipment payments while operations consumed $(73.8) million of cash. It raised $1.028 billion in net equity proceeds over that period, including $538.8 million net under a $600 million at-the-market programme opened in May 2026.

How much has AAOI diluted shareholders?

Shares outstanding rose from 75.0 million at 31 December 2025 to 84.4 million at 30 June 2026, an increase of roughly 12.5% in six months. The company’s own Q3 2026 guidance assumes a diluted share count of about 92.8 million, so the trend continued past the half-year mark.

Does Amazon have to buy $4 billion of AAOI products?

No. Amazon holds a warrant issued in March 2025 for up to 7,945,399 shares at $23.6956. Only 1,324,233 shares were exercisable at issuance. The remaining 6,621,166 vest only if Amazon and its affiliates purchase $4 billion of AOI product over the warrant’s ten-year term. It is an incentive tied to a milestone, not a commitment to purchase.

Is the US actually banning Chinese optical transceivers?

Not yet. Reuters reported in August 2026 that the FCC was drafting a rule to ban imports of new Chinese optical transceiver models for AI data centres. It has not been formally proposed or approved, reporting noted the agency could still modify or shelve it, and existing installed equipment would likely be exempt. Treat it as optionality rather than as a fact in the investment case.

What is co-packaged optics and why does it matter for AAOI?

Co-packaged optics places the optical engine directly alongside the switch chip instead of in a separate pluggable module, cutting power consumption and shortening the electrical path. For AOI it shifts the product mix toward lasers and light sources, where management targets gross margins of roughly 55% to 65%, well above what the transceiver business earns today.

What is the single most important number to watch in the next result?

Gross margin. Revenue growth is already assured by the capacity ramp and the order book. Whether that revenue becomes profit, fast enough to outpace the rising share count, is the entire investment question, and gross margin is where it shows up first.

—

Financial Disclaimer:

This article is for educational and informational purposes only and should not be considered financial or investment advice. Investing involves risk, including the possible loss of capital. Always conduct your own research and consult a qualified financial adviser before making investment decisions.