AI Optics 2026: The Memory-to-Fiber Rotation, Read Through $CIEN, $LITE and $POET

01 What lit the optical sector

The draft ban on Chinese optical modules

On August 4, 2026, reports circulated that the Trump administration and the Federal Communications Commission are drafting a ban on US imports of Chinese optical transceivers for data centers, the components that route data over fiber inside AI compute clusters. The stated aim is to finalize and enforce the rule before year end.

The reaction was immediate and concentrated on Western suppliers: Applied Optoelectronics gained 17%, Coherent 11%, Lumentum 6% in the same session.

The Applied Optoelectronics print

On August 6 the theme got a second push on firmer ground: Applied Optoelectronics reported revenue of $191.9 million, up 86.3% from $103.0 million a year earlier, returning to profit on an adjusted basis. In the next session, on August 7, Coherent closed +13.4% and Lumentum +6.2%, pulling the whole photonics basket along.

The rotation thesis: memory into optics

The dominant weekend narrative is a sector rotation. The argument, attributed to a Citrini analyst and repeated by numerous users, runs as follows: Nvidia would use a smaller share of HBM memory on the Rubin Ultra generation while leaning more heavily on optical networking, and memory pricing is seen near a cycle peak. Hence the shorthand circulating everywhere: sell memory, buy optics.

This is a short-term positioning thesis and should be treated as such: the reading of market participants, not established fact. The same boards carry the opposite argument, that recent weakness in memory names stems from forced selling by leveraged Korean ETFs, a technical factor rather than a fundamental one.

02 The sector on the tape and on the numbers

Prices and sentiment, August 7, 2026 closes

Two things deserve attention. First: POET carries the highest watcher count in the group, 29,135, above AAOI and more than four times Ciena’s, despite being by far the smallest company by revenue. Second: Ciena is the least-discussed name, with normal message volume and just 6,841 watchers, despite being second in the group by quarterly revenue. In this sector, retail attention and industrial scale run in opposite directions.

The fundamental comparison: who actually sells what

The table above measures how prices moved and how much they are discussed. It says little about what sits underneath. The comparison below places side by side the last quarter each company actually reported.

The order of magnitude. The smallest peer in the group, Applied Optoelectronics, books 381 times POET’s quarterly revenue in a single quarter. Ciena books roughly 3,100 times as much, Coherent about 3,600. This is not a difference in business quality: it is a difference in stage.

Growth is not the differentiator. The fastest grower is Lumentum at +90.1%, followed by Applied Optoelectronics at +86.3% and Ciena at +39.5%. Full-acceleration numbers, but achieved on already substantial revenue bases. The theme is not rewarding those who promise growth: it is rewarding those already booking it.

Profitability splits the group. Lumentum runs a 47.9% non-GAAP gross margin and $2.37 of earnings per share. Ciena sits at 44.9% adjusted with $1.64 and a 19.5% adjusted operating margin. Coherent at 39.6% with $1.41. Applied Optoelectronics has only just returned to adjusted profit at $0.06, remains at a GAAP loss of ($0.28), and carries a 29.8% gross margin contracting from 30.4% a year earlier. Even among the names rising together, earnings quality varies widely.

03 $CIEN — the one with the numbers that is not being paid for them

Ciena is the most interesting name in the group precisely because it is the least discussed. It does not make transceivers: it builds optical systems and network equipment, which places it at a different level of the stack from Coherent, Lumentum or Applied Optoelectronics. It was not named among the direct beneficiaries of the draft Chinese-module ban, and indeed it did not join the sector jump on August 4.

The fiscal second quarter, reported June 4, 2026

• Revenue $1.57 billion, up 39.5% from $1.126 billion a year earlier
• Adjusted gross margin 44.9%, up from 41.0%; GAAP 44.0% from 40.2%
• Adjusted operating margin 19.5%, up from 8.2%; GAAP 15.1% from 2.9%
• Adjusted diluted EPS $1.64, up from $0.42; GAAP $1.49 from $0.06
• Cash and equivalents $1.045 billion, total assets $6.039 billion
• Order backlog $7.7 billion, up sequentially

The composition matters as much. Cloud providers accounted for one third of quarterly revenue and direct cloud-customer revenue grew 70% year over year. Optical networking rose more than 40%, but the fastest-accelerating segment was routing and switching at +88%. Service provider revenue grew 28%, and revenue from Indian service providers more than doubled. The company also received its first multi-rail order for the RLS HyperRail photonic system from an unnamed hyperscaler.

Guidance was raised: $6.3 billion ±$100 million for full-year 2026, roughly 32% growth at the midpoint, with adjusted gross margin of 44.5% to 45% and adjusted operating margin of 19% ±50 basis points. For the third quarter the guide is $1.625 billion ±$50 million at a 19% to 20% adjusted operating margin.

Why the stock fell anyway

Here sits the real informational content of the Ciena case. On June 5, the day after that print, the stock fell nearly 14%, closing down 13.66%. Three reasons are reported, and none concerns the quality of the results.

1. Valuation. The stock came in from a gain of over 600% in a year and traded at a forward price/earnings ratio of around 90 times before the release.
2. Expectations already embedded. The beat was marginal against what the market had already priced, a textbook buy the rumor, sell the news reaction.
3. Customer concentration. In the quarter two customers represented 34% of revenue. It is a disclosed figure that weighs on perceived growth quality, alongside supply-chain bottlenecks cited as a constraint on expansion.

04 $LITE — the highest margin and fastest growth

Lumentum is the mirror image of Ciena: excellent numbers and a price that follows them. In fiscal third quarter 2026, reported May 5, it posted revenue of $808.4 million, up 90.1% year over year and 21.5% sequentially. That is the fastest growth in the entire basket.

Profitability is the best of the group: non-GAAP gross margin of 47.9%, GAAP 44.2%, with non-GAAP diluted EPS of $2.37 against $1.50 on a GAAP basis. Revenue splits into components at $533.3 million, 66% of the total, and systems at $275.1 million, 34%. Management attributed margin expansion to operational rigor, pricing discipline and product mix, citing strength in laser chips and scale-across components such as pump lasers.

Fiscal fourth quarter guidance is $960 million to $1.01 billion of revenue, with non-GAAP operating margin of 35% to 36% and non-GAAP diluted EPS of $2.85 to $3.05. If hit, it would take the company above a billion dollars of quarterly revenue.

05 $POET — the technology option and the capital bill

POET Technologies plays a different game from all of the above. It is not a manufacturer riding the cycle: it is a company trying to turn a photonic integration platform into volume production. We have already published a dedicated explainer on how the Optical Interposer works and where POET sits in the optical stack: what matters here is the financial side.

The first quarter 2026 numbers

On May 14, 2026, POET reported revenue of $503,389, a net loss of $12.34 million or $0.08 per share, research and development of $4.50 million, wages and benefits of $4.05 million, stock-based compensation of $3.49 million. Operating cash flow was negative $8.8 million, which annualized gives roughly $35 million a year of burn on a rising cost base. Cash and short-term investments at March 31: $429.14 million, before the $400 million received in May.

The most direct comparison: stock-based compensation alone for the quarter is worth about seven times total quarterly revenue. The cost structure is that of an industrial company under construction, the revenue is still that of a pre-commercial phase.

Capital: $830 million raised, and at what price

Between May 2025 and May 2026, POET raised $830 million in gross proceeds across six equity financings. The largest is the one that matters for understanding the present:

• May 18, 2026: registered direct offering of $400,000,020 with a single institutional buyer
• 19,047,620 shares issued at a combined unit price of $21.00
• 19,047,620 accompanying warrants, struck at $26.25, a 25% premium to the offering price, with a three-year term

Shares outstanding went from 76,507,157 on January 1, 2025 to 172,595,406 on May 18, 2026: a 126% increase in roughly seventeen months. That is the cost of the $830 million.

The stock, meanwhile, closed at $8.91 against the $21.00 offering price of May: -57.6% in under three months. Against the $20.57 closing high of May 14 the decline is 56.7%.

When the second quarter arrives

Four lines will carry information: the June 30 cash balance, whether anything has been recognized against the two open orders, the share count on the cover of the interim statements, and additions to property and equipment, $2.43 million in the first quarter against a stated plan of roughly $50 million of capex for the second half. That is where a tenfold capacity expansion would first become visible.

06 The calendar: three dates, three degrees of certainty

Across four sessions the sector faces two tests: POET from the pre-revenue side and Coherent from the consolidated-numbers side. The second carries far more weight for whether the theme holds.

07 Risks and red flags

On the theme overall

• Regulatory risk. A large part of the rally rests on an unenacted draft. A narrowing would remove support from the entire basket.
• Rotation risk. If the flow from memory into optics exhausts itself, the factor supporting the sector disappears regardless of fundamentals.
• Valuations. The Ciena episode of June 5 shows that here a quarter growing 39.5% with raised guidance can still produce a 14% decline.

On $CIEN

• Customer concentration: two customers at 34% of revenue in fiscal Q2.
• Supply-chain bottlenecks, cited as a constraint on meeting demand.
• Valuation sensitivity: the June precedent is documented and recent.

On $LITE

• An increasingly high comparison base: after 90.1% growth, year-over-year comparisons get progressively harder.
• Deteriorating sentiment despite the best numbers in the group, with a delta of -22.22.

On $POET

• Structural dilution: +126% shares in seventeen months, and roughly 261 million shares on full exercise.
• May warrants deeply out of the money: struck at $26.25 against an $8.91 price.
• Revenue still nominal against a cost structure orders of magnitude larger.
• Industrial execution risk: one million units a month by end-2027 requires roughly ten times current capacity.
• No guidance and no earnings call, with limited independent analyst coverage.
• Elevated short interest: 30.11% of float per Finviz data as of August 7, 2026.

08 Possible scenarios

What follows describes possible configurations. It is neither a forecast nor an operational indication.

The third possibility: the rotation reverses. Should the memory-cycle-peak thesis prove premature, the flow that supported optics in this phase could turn back. It would be a positioning move rather than a judgment on individual fundamentals, but it would affect prices all the same.

09 What we could not verify

• The $5.5 billion Lumilens valuation circulating on Stocktwits: we did not find it in company releases or filings. The warrant granted to Lumilens is documented, up to 22,921,408 shares at $8.25 over nine years, of which 2,292,140 immediately exercisable.
• Applied Optoelectronics backlog and order data cited in social commentary, in particular 800G demand booked out to mid-2027: not verified against the official release, which discloses no backlog figures.
• The identity of the investor in POET’s May offering: company releases describe it only as a single institutional investor.
• The absolute split of Ciena revenue between cloud and service provider: the release gives growth percentages and the one-third share of total revenue, but not the precise breakdown.

10 Merlintrader bottom line

The rotation into optics is real and stands on two legs: a regulatory one, still at draft stage, and an industrial one, supported by the accounts of companies that already produce and sell. Only the second is verifiable today, and the numbers describe it precisely: Ciena at $1.57 billion with a 19.5% adjusted operating margin and $7.7 billion of backlog; Lumentum at $808 million with a 47.9% gross margin and guidance taking it toward a billion a quarter; Coherent at $1.81 billion awaiting its full-year print.

Ciena is the case that deserves more attention than it gets: it has the growth, the margin, the backlog and an announced date, and it is the least-discussed name in the group. Its problem was not the quarter, it was the price it arrived at.

POET participates in the same theme from a position that differs in kind, not in degree. It has ample liquidity, a platform with a coherent engineering argument and stated industrial targets; it also has $503 thousand of quarterly revenue, no guidance, no earnings call, and half a billion dollars of warrants that remain paper at these prices. The interim statements due by August 14 are the first document that can move anything on that front.

Related research on Merlintrader

• POET Technologies ($POET) Stock Hub 2026 — cash, capital structure, funded orders versus announcements, catalyst table
• Why AI Needs Optics: the optical interconnect stack explained — the five layers and where POET sits
• Three different ways to play the AI photonics race: $POET, $COHR and $AAOI
• Dilution, ATM offerings, PIPE deals and reverse splits: a practical guide