revoAIution

Can $SIVE and $AAOI turn a ~$100K account into $1M+ by 2030? I’m going to track it and see what happens. I started a separate account holding only: • 500 shares of $AAOI • 5,000 shares of $SIVE The idea is simple. I believe AI infrastructure is still in the early stages of a massive buildout, and optical connectivity could be one of the biggest areas to benefit. In the article I go through why I’m weighted more toward AAOI today, why I still believe in the $SIVE thesis, how I’m thinking about the upside, the risks involved, and what this account could become if the thesis is right. This is a high-risk experiment and I’m sharing the journey as it happens. Free Substack article below revoaiution.substack.com/p/t…

$AAOI + $SIVE Weekly | Week 4 | Nov 2-6 A brutal week: $AAOI beat but pushed out capacity, $SIVE fell with chips, and Lumentum signed $5B+ of CPO demand. WHAT HAPPENED $AAOI 1. Dropped 21% on Thursday before it even reported, as optics names sold off together after Coherent's results. Slid again after hours on the guide, then bounced 7% Friday to $50.13. Down 27% on the week. 2. Q3 came in at a record $269.9M, up 128% and inside the $255-290M guide, the sixth straight record quarter. Data center revenue jumped 254% to $155.6M and CATV hit a record $110.9M. GAAP profit returned at $6.3M, though other income did the heavy lifting there. Non-GAAP gross margin of 28.7% came in under the 29-30.5% guide. 3. Q4 is guided to $385-420M, below the ~$427M the Street wanted, with data center at $270-300M and 1.6T alone at $50-70M. Management said a big customer moved from 800G to 1.6T faster than expected and DSP supply is the main thing capping 1.6T right now. 4. Capacity is at ~330k units a month, up from ~200k at Q2. The year-end target dropped to 500k+ from 650k because of construction and electrical power delays in Texas. The end-2027 target went up to 1M+ from 930k. Sugar Land starts initial operations in December. Management now frames the ~$470M a month data center revenue goal around the end of 2027, versus mid-2027 on the Q1 call. 5. $AAOI signed a long-term deal with an existing hyperscaler for at least 10% of its 1.6T needs from June 2026 through May 2029, worth a potential $1-1.5B. The CEO said a second deal is in final stages and could be five times bigger or more, covering 800G and lasers. It's not signed yet. 6. Funding for the buildout came from a new $200M bank facility signed Oct 30, $170M in term loans plus a $30M revolver, with no new share sale announced. 7. Analysts cut targets: Rosenblatt to $115 from $220 (Buy), Needham to $100 from $170 (Buy), B. Riley to $45 from $92 (Neutral). $SIVE 1. Fell from 20.90 to 16.20 SEK, down 22.5%. Thursday alone was down 19% before a 7% bounce Friday. No company news. It moved with chips and optics. 2. CEO Vathulya did an EE Times podcast. Per notes from listeners: Jabil is still in beta builds, the LiDAR customer ramps out of Glasgow starting this quarter, and $SIVE has shifted toward a production manufacturing model with about one unit of Glasgow capacity for every 2-2.5 at partners. He also said lasers have to keep going up in power as speeds rise, and that $SIVE is building for pluggables, LPO, LRO and NPO alongside CPO. That spreads the bet beyond the one product type everyone is waiting on. 3. Q3 report Nov 26. SECTOR READ-THROUGHS 1. Lumentum posted $1.26B in revenue, guided next quarter to $1.575-1.625B with operating margin of 45-46%, and said it landed its second major NPO/CPO win, backed by more than $5B in purchase commitments. That's real money committed to lasers for co-packaged and near-packaged optics, the exact lane $SIVE is building for with its CW lasers and arrays. 2. Marvell made the Lumentum deal official: all stock, 4.5 Marvell shares per Lumentum share, about $1,132 a share and $115B in enterprise value, closing in H1 2028. A chip company paying that much for a laser maker puts a big price tag on laser capacity, which is what both $AAOI and $SIVE make. 3. Coherent posted revenue up 50% to $2.37B, data center and comms up 79% to $1.95B, and guided fiscal 2027 growth above 60%. The stock still fell about 14% Thursday and dragged $AAOI and the rest of optics with it. The demand numbers keep beating while the stocks keep falling, so the selloff is about sentiment and funding more than orders. 4. Democrats won both chambers of Congress, including a Senate seat and the governor's race in Texas. The Dow jumped 3% Wednesday as oil fell, but chips and AI infrastructure sold off on worries about tougher data center rules and costs. Texas is one of the biggest AI build states, so this touches both names through their customers. Policy can change where and how fast clusters go up, but the budgets come from Amazon, Microsoft, Google and Meta. 5. The hardware money is still flowing. AMD's purchase commitments hit $48.8B, four times the end-2025 level, with $33.4B due by end of 2027. Goldman and Apollo started marketing SpaceX's $40B financing for about $47.3B of NVIDIA Rubin gear. Alphabet committed $15B to Malaysia data centers for 2027-28. Korea's October exports jumped 94%, with chips up 259% to a record $56B. Big Tech buybacks fell to $4.4B in Q3, the lowest since 2017, because the cash is going into capex instead. All of that becomes orders for parts, optics included, through 2027. COMING UP 1. Nov 17: NVIDIA Q3 2. Nov 26: $SIVE Q3 3. Any week now: $AAOI's second hyperscaler deal, per the CEO PRICE TARGETS (end of year) $AAOI 2028: $883 (~$80B) $SIVE 2028: $98.11 (~$35B) Changes this week: Removed end of 26,27 price targets. Focusing on a 2+ year trade

Retail is bailing right at the start of the AI buildout Retail just had its biggest one-day exit from US stocks since 2020. We’re at the beginning of the AI revolution, and this is happening. I’m break even or slightly down on a lot of my AI trades right now too, so I get the frustration. But this trade was never get rich quick. The buildout runs for years, and the people who make money on it are the ones still holding when the revenue actually shows up. Same with the $AAOI print. Q3 revenue came in at $269.9M, their sixth straight record quarter. Were there delays? Of course. They’re scaling 800G and 1.6T production faster than this industry has ever had to. Nothing goes perfectly. What I care about is that management is being upfront about it now. Given their history, that’s reassuring. I’d rather hear about a delay on the call than find out a quarter later. Selling the start of a multi-year buildout because of one ugly stretch is how retail keeps getting shaken out of the best trades.

$AAOI reports tonight $AAOI reports Q3 after the close today, and this is the one that tells us whether the October and now November sell off has anything to do with the business. What a wild sell off. But earnings calls are what keep the panic at bay. Today we find out if the sell off is warranted or not. What I want to see: 1. Capacity. They exited Q2 around 200k units a month of 800G and 1.6T and targeted 650k+. I want to see it expanded and hear whether Sugar Land started production. 2. The 2027 target. On the Q1 call, CFO Stefan Murry laid out about $471M a month of data center transceiver revenue by mid-2027. I want to hear that path reaffirmed. 3. Customers. Management said demand runs ahead of capacity through mid-2027. I want to hear the hyperscalers are still at the plate and still switching from copper to light. 4. The numbers. Q3 was guided to $255-290M. Then gross margin and the Q4 guide. Lumentum reports tonight too, so we get two reads on the laser side at once. If they hit, the sell off was sentiment. If they miss on capacity or the 2027 path, the market had a point. Either way, today we find out if the market is right or not.

It’s looking bad right now, but $SIVE doesn’t just serve the States. $SIVE is down to around $1.81 USD from a high near $11. AI sentiment has turned, and some Democrats are pushing a national pause on AI data center construction. I don’t think a pause actually happens. It might look good publicly, but it would be a disaster for the US. It would be Canada all over again: a country sitting on everything it needed to become a powerhouse, with a government that never used it. A huge opportunity, fumbled. And if the US does fumble it, someone else takes the wheel. AI doesn’t stop because one country hits the brakes. The buildout just moves. That’s why I like where $SIVE sits. It’s a Swedish company with a fab in Scotland, selling lasers worldwide. It’s already working with China-based O-Net on an external laser source module. Whoever wins the AI race needs lasers to build it. US, Europe, Asia, it doesn’t matter. Every data center needs light to move data. I’m not betting on one country winning. I’m betting on the buildout happening somewhere, and $SIVE selling into it wherever that is.

Ok so little midweek update for $AAOI and $SIVE, because two big things happened this week and it feels like the AI trade is over... it's not. First, Marvell is buying Lumentum. That's a $115B price tag on a laser and transceiver company. $AAOI does lasers and transceivers too and sits around $6B. $SIVE makes CW lasers in its own InP fab and sits under $1B. Different size, different stage, but the market now has a number for what laser capacity is worth. It also shrinks the list of independent laser suppliers. Once Lumentum is inside Marvell, a lot of buyers won't want to depend on a chip company for their light. My read is that helps $SIVE. For $AAOI, the thing to watch is DSPs, since Marvell sells them to transceiver makers and now competes with some of its own customers. Second, Democrats won the governor's race in Texas. Texas has been one of the biggest places to build AI data centers in the US, and plenty of what $AAOI and $SIVE ship probably ends up in clusters there. So this one does touch both names, through their customers. Grid rules, power costs and data center tax breaks are all set at the state level, and a new governor can push on those. If Texas gets tougher or slower, some projects shift timelines or move to other states. But the demand doesn't come from Austin. It comes from Amazon, Microsoft, Google and Meta, and they just told us 2027 capex goes higher. Wherever those clusters land, they still need transceivers and lasers. Texas policy decides where and how fast. It doesn't decide whether. $AAOI also builds in Texas. Sugar Land, Pearland and the Houston buildings are factories, already under way, with state grant money already awarded. I'm not changing my price targets yet. I still think we hit our targets end of 2030 even though right now it feels like it's over.

$AAOI + $SIVE Monthly | October recap, November ahead October brought AI bubble talk and a nasty selloff, and Big Tech answered by raising AI spending again. That's the setup for $AAOI and $SIVE going into November. TOP 5 FROM OCTOBER 1. The spenders raised the bar again. Amazon kept 2026 capex near $220B, said 2027 goes meaningfully higher, and expects AWS to grow over 40% in both 2027 and 2028 on a $578B backlog. Alphabet held 2026 at $205B and promised a big step up in 2027. Meta raised $25B in new stock to keep building. JPMorgan now sees AI data center capex up 85% in 2026 and 70% in 2027, from about $800B to $1.5T. Every one of those dollars turns into racks wired with optics, which is $AAOI's transceiver business and the CPO wave $SIVE's lasers are built for. 2. $SIVE's LiDAR program went to production. The strategic LiDAR customer placed its first production orders on Oct 20, with shipments starting this quarter and rolling orders into 2027. Shareholders also approved EY as auditor for the US listing, and Sivers is hiring a Silicon Valley head of IR ahead of the planned H1 2027 Nasdaq listing. On the defense side, ALL SPACE landed a Viasat Government order for the Hydra MAX terminals that run on Ka-band beamforming chips like the ones $SIVE supplies to ALL SPACE. 3. Optics got louder. Nvidia is targeting a 15x expansion in optics capacity by 2027 for CPO. At OCP, Ayar Labs ran a live CPO compute system on its SuperNova light source, the product $SIVE's lasers power. Super Micro beat its own revenue guide by about 27% at the midpoint on AI demand. More AI hardware shipping means more optics, and $AAOI is adding capacity to catch it. 4. Lasers became the bottleneck story. Bloomberg reported Marvell is in advanced talks to buy Lumentum, which Bloomberg says is sold out into early 2029. AXT said its 2026 indium phosphide revenue opportunity has more than quadrupled, and Morgan Stanley sees InP substrates, EML and CW lasers as the main bottlenecks. $AAOI makes its own lasers and locked in long-term InP supply. $SIVE makes CW lasers in its own InP fab in Glasgow. Both sit on the scarce side of the chain. 5. The stocks still got hit. Hyperscalers are paying for the buildout with debt and new stock, the Fed hiked in September and stayed hawkish in October, and chip stocks sold off hard on bubble fears. $AAOI finished its third share sale of the year, and WIN Semi said CW laser volume is a 2028 story with small revenue in 2027. The selling was real, and so is the dilution. 5 THINGS FOR NOVEMBER 1. Tue Nov 3: US midterms and Super Micro's full results. One clears a big macro overhang, and the other shows how much AI hardware actually shipped. 2. Thu Nov 5: $AAOI Q3 after the close, against a $255-290M guide. What I'm watching: Sugar Land production, capacity vs the 650k units/mo year-end target, gross margin, the 1.6T ramp and the Q4 guide. 3. Thu Nov 5: Lumentum reports the same night, and a Marvell deal could land any day. It's a read on laser pricing, external laser source demand, and who ends up controlling laser supply, all of which matter for $SIVE. 4. Tue Nov 17: NVIDIA reports Q3 after guiding $108B, with an early call for about 70% revenue growth next fiscal year limited by supply. Any update on Vera Rubin timing and optical scale-up flows straight down to $AAOI and $SIVE. 5. Thu Nov 26: $SIVE Q3. The list: is the Q4 revenue going to show up with first LiDAR shipments, progress at WIN and the second foundry partner, Jabil beta builds, Glasgow, cash, new pluggables and NPO customers, and Nasdaq timing, WHERE I STAND Nothing in October changed my view of this trade. The stocks sold off and the bubble talk got louder, while the companies writing the checks raised their budgets again. Amazon, Alphabet and Meta all point to more spending in 2027, and JPMorgan's numbers say the same. That spending has to run on optics, and lasers are already the part in short supply. For $AAOI, it shows up as transceiver orders for capacity it's still building. Nov 5 is about execution. For $SIVE, the timeline is longer. LiDAR revenue starts this quarter, pluggables come in 2027, and CPO lasers scale around 2028, which lines up with what WIN said. The selloff is about how the buildout gets financed and about sentiment. The demand side didn't break.

$AAOI + $SIVE Weekly | Week 3 | Oct 26-30 Absolute insane week with a bunch of headlines. Marvell went after Lumentum's lasers, and $AAOI and $SIVE still gave back most of their gains in Friday's AI selloff. WHAT HAPPENED $AAOI 1. Ran from $67.51 to $82.22 by Wednesday's close, up 22%, then gave most of it back: down 3.7% Thursday and 12.8% Friday as the Amazon and Google results revived worries about AI spending and optics names sold off together. Still finished the week up 2.3%. 2. Needham trimmed its $AAOI target to $170 from $190 and kept its Buy rating. 3. Q3 lands Thursday, Nov 5 after the close. That's when we find out if Sugar Land started production and how close capacity to units/mo target. $SIVE 1. Climbed from 20.64 to 23.12 SEK by Thursday, up 12%, then dropped 9.6% in Friday's shortened Stockholm session to 20.90. Up 1.3% on the week. 2. No company news. The next real update is the Q3 report on Nov 26. SECTOR READ-THROUGHS 1. Big Tech's message was spend more, and the market didn't like the bill. Amazon's AWS grew 39% to $46.0B, and Amazon guided AWS growth above 40% in both 2027 and 2028, holds a $578B backlog, kept 2026 capex near $220B and said 2027 goes meaningfully higher. Microsoft's Azure hit $33.2B and Google Cloud $30.7B, and Alphabet held its 2026 capex guide at $205B with a big step up promised for 2027. The catch: Microsoft and Alphabet both spent less than expected in the quarter, about $39B vs ~$46B and $51.4B vs $55.8B (Microsoft's July accounting change lowers its reported number), and Amazon's EPS missed. Every one of these clouds runs on optics, so bigger 2027 budgets are the demand $AAOI is building for and the CPO wave $SIVE is waiting on. The market sold the spenders and their suppliers anyway. 2. Bloomberg reported Wednesday that Marvell is in advanced talks to buy Lumentum, with a deal possibly announced as soon as this coming week. Neither company has confirmed it. Lumentum jumped about 11% after hours toward ~$91B, then both fell 8-10% Friday with the rest of optics. A $220B chip company reaching for a laser maker that Bloomberg says is sold out into early 2029 shows where the scarcity is. For $SIVE, if the laser leader Morgan Stanley grouped it with ends up feeding Marvell first, other CPO builders need more laser sources, which is my read and unconfirmed. For $AAOI, it makes its own lasers and transceivers, the combination Marvell wants, but Marvell owning Lumentum would also turn a DSP supplier into a competitor for module makers. 3. WIN Semi, $SIVE's CW laser foundry partner, reported Q3 revenue up 31% with gross margin at 37%, and infrastructure plus optical made up more than half its revenue for the first time in a third quarter. The less exciting part: WIN said its optical growth came from receivers, and that its CW laser and EML products are through customer development but still in reliability testing, with small revenue in 2027 and volume production in 2028. That lines up with $SIVE's own timeline of Jabil production orders in H1 2027 and a ramp in H2 2027, so real laser volume lands late 2027 into 2028. WIN is also pushing toward 200-400mW high-power lasers, the class CPO needs. The Nov 26 Q3 call is the next chance for $SIVE management to say where the WIN program stands. 4. The InP squeeze keeps showing up. AXT, a major indium phosphide substrate supplier, posted Q3 revenue of $73.4M, up from $47.6M last quarter and $28.0M a year ago, with gross margin at 56%, and said its 2026 InP revenue opportunity has more than quadrupled. Taiwan's VPEC told UDN that InP substrates are in serious short supply, and Morgan Stanley sees InP substrates, EML lasers and CW lasers as the main bottlenecks. $AAOI locked in long-term InP substrate supply this year and makes its own lasers. $SIVE runs its own InP fab in Glasgow and makes CW lasers. Both sit on the scarce side of the chain. 5. Anthropic's reported S-1 shows H1 2026 revenue of $16.3B, up from $1.16B a year earlier, and compute commitments running out to 2045: about $161B in Broadcom TPU leases, $120.5B with SpaceXAI, $114.5B with Google, $110B with Amazon and $44.9B with Microsoft, plus roughly $245B more signed after June 30. That's around $800B of contracted compute from one customer, and every cluster in it gets wired with optics. Amazon's slice matters most for $AAOI, since Amazon holds warrants on $AAOI shares. 6. The money for the buildout showed up this week, at a price. SpaceX signed a binding commitment for a $40B GPU lease deal through an Apollo-managed entity, with deliveries from H1 2027. Meta upsized its stock sale to $25B at $625 a share, with NVIDIA in for $3B and AMD $1B. IREN borrowed $363M against GPUs already earning under multi-year contracts, and UBS said it doesn't see capital getting scarce in 2027. Chip suppliers funding their own customers keeps orders moving toward optics names like $AAOI and $SIVE, even as all that dilution and debt weighs on the stocks. 7. SK Hynix beat with $78.9B in revenue and guided Q4 DRAM shipments up 25% QoQ. It now has 5-year supply deals with 14 large customers, backed by deposits, covering more than half its capacity in both 2027 and 2028. More DRAM shipping should ease the memory shortage that held up one $AAOI customer's 100G switch builds, and buyers locking in parts years out shows how hard everyone is planning to build. 8. The Fed held at 3.75-4.00% Wednesday in a 10-2 vote, with two members wanting a hike, and markets still lean toward a December hike after September's increase. Friday the Nasdaq fell 2.86%, its worst day since June, on fresh doubts that AI spending pays off. Hyperscalers are funding the buildout with more debt and stock, semis aren't getting rewarded for it, and leveraged dip buyers poured a record $9B into SOXL in two weeks. That mix makes small optics names like $AAOI and $SIVE swing hard both ways, and this week showed it. NEXT WEEK 1. Tue Nov 3: US midterm elections, Super Micro full results 2. Thu Nov 5: $AAOI Q3 after the close 3. Any day: a possible Marvell-Lumentum announcement 4. Coming up: $SIVE Q3 Nov 26 PRICE TARGETS (end of year) $AAOI 2026: $150-240 (~$13.6-21.7B mkt cap) 2027: $441 (~$40B) 2028: $883 (~$80B) $SIVE 2026: $3-5 (~$1.1-1.8B mkt cap) 2027: $22.43 (~$8B) 2028: $98.11 (~$35B) Changes this week: none