A lot of discussion of Taiwan seems to ignore Taiwan's potential strategic moves. These include:
- Nuclear programs. Until the late 80s, Taiwan was secretly a nuclear power in all but name. They shut down their official nuclear program based on US pressure. But nuclear weapons aren't that hard, and anyone who can build TSMC chip fabs is entirely capable of building both fision and fusion weapons, given access to uranium. And Taiwan has a history of hiding the necessary technological development as dual use programs.
- Conventional strategic weapons. Taiwan manufactures both domestic fighter jets and multiple tiers of cruise missiles. Looking at the publicly-known range numbers, it's pretty clear that they could hit strategic targets like the Three Gorges Dam with conventional bunker-buster warheads unless intercepted by Chinese air defense.
- Taiwan has seen this potential conflict coming for decades, and they have historically kept strategic programs like their nuclear program secret (perhaps to minimize escalation). Since 2016, it's been clear that the US is either an unreliable strategic partner for democratic nations, and the alarms have been blaring since last summer. And it has been clear since about 2002 that any non-nuclear nation is the plaything of the big powers, and outside of NATO, only deeply foolish nations wouldn't be investing in at least the precursors to nuclear capability. Proliferation sucks, but it's also the only way to protect your country from invasion.
So whenever people speak of the inevitable invasion of Taiwan by China, I'm always looking to see their analysis of Taiwan's counter-moves. What's their timeline for Taiwan having fision/fusion weapons, should Taiwan choose to pursue that again? What's their analysis of Taiwan's conventional strike capability against strategic targets? Maybe it's self-evident to actual experts that Taiwan has no viable options here. But I rarely see any discussion of whether Taiwan could escalate into a Mutually Assured Destruction dynamic, which is confusing when we're talking about a former nuclear power (in all but name) that continues to invest heavily in cruise missiles that can reach most key targets in China.
So I'm prepared to be convinced by experts here! But based on just public knowledge, I can't rule out the possibility that Taiwan has strong counter-moves, and a past ability to prepare in secret. So a lot of this comes down to expert knowledge of the IAEA inspections, where all Taiwan's uranium purchases went, the political likelihood of Taiwan's current leadership pursuing a program like this, etc. The US appears to have been officially "surprised" by Taiwan's nuclear capabilities at least once before, and maybe there's no way that could actually happen again. But I'd love to see the expert argument!
I sincerely hope Taiwan has its act together and has plans to offer a credible deterrent to China. What I've read about Taiwan's military, though, often does not really instill much confidence. I can't find great sources at the moment but this goes over some of the issues -- personnel shortages, poor training, lack of investment, and a focus on conventional systems like large warships at the expense of asymmetric capabilities. I know the current government is working on improving these issues, and they might have various secret plans, but from the level of competence they've generally displayed I'm not very optimistic.
1&3: Even if Taiwan maintains its non-nuclear status, Beijing's intent to wage a unification war is increasingly overshadowing concerns about economic sanctions and casualties. Should Taipei attempt to acquire nuclear weapons again, it would trigger tensions far exceeding those of the North Korean nuclear crisis or the THAAD crisis, making war highly probable.
Acquiring nuclear weapons is fundamentally different from gaining the capability to deploy them. The advantage of nuclear terrorism gained through a small number of primitive fission devices would not secure victory for Taiwan, just as Iraq did not win the Gulf War through its chemical weapons advantage. If these devices are not destroyed, captured, or neutralized early in the conflict, their sole utility would be for scorched-earth tactics—but Taipei's leadership is unlikely to descend into madness.
The United States is unwilling to engage in nuclear warfare. Therefore, should Taipei's leadership exhibit overtly irrational behavior, Washington would likely refuse assistance, leaving Taiwan incapable of prevailing alone.
Taiwan cannot independently manufacture all equipment required for TSMC chip factories; its lithography machines and other apparatus rely on imports. Should Taiwan attempt to import centrifuges after its nuclear program is exposed, it might have to resort to submarine transport.
2: As the Three Gorges Dam is a gravity dam. Most conventional missiles cannot destroy it at an acceptable cost. To demolish it would require shattering hundreds of millions of tons of reinforced concrete.
China possesses robust air defense and anti-missile systems, while Taiwan's missile technology remains at the PLA's 2000s level. Even if the Taipei regime planned to strike mainland China before its launch platforms were destroyed, the civilian targets it could effectively attack would primarily be urban clusters along the Fujian coast.
1Thank you for your detailed response!
This has given me several hypotheses that seem worth further investigation. I need to go look at cruise missile specs again, at the very least.
But I rarely see any discussion of whether Taiwan could escalate into a Mutually Assured Destruction dynamic
Not at all an expert either. But MAD doesn't just assume nukes but also assumes that each side is able to launch on warning/under attack, or to launch a second strike. The former is harder, compared to US-Russia, because of the physical proximity. The latter disadvantages Taiwan because of its smaller land mass and much more limited navy. So maybe experts have very basic reasons to question the viability of MAD?
No one except outright tyrants would bomb civilian infrastructure or use nuclear weapons, subjecting millions of people to suffering. But there is a far more effective defensive weapon, striking in its cheapness — fiber-optic FPV kamikaze drones. They can be assembled from items sold in ordinary stores for a few thousand dollars, adding some tape and a shaped-charge warhead. You can find videos showing this deadly toy flying into the open window of a moving car or into a tank’s air intake, rendering a meter-thick frontal armor meaningless. No electronic warfare can interfere with its effective operation, and its low cost allows such drones to be used even against individual soldiers.
https://www.youtube.com/watch?v=JcB4dR_8xPo
And you see a perfect picture right up to the very end, sitting somewhere 1,000 km from the front line in FPV goggles, while ordinary soldiers simply load new drones for you, and your commander monitors the overall battlefield situation from a reconnaissance drone.
The war of the future looks nothing like nuclear strikes or the destruction of dams; it is an endless web of fiber optics, with a deadly charge at the end of every strand.
https://www.youtube.com/watch?v=jr7M-AmrvT4
In this war, defense surpasses offense: if you have hundreds of thousands of such drones, your defense is impenetrable, and only the ambitions of mad politicians demand the continuation of the show.
Developments in AI are very important here, because they will allow such a drone to fly without any fiber-optic cable at all, simply by recognizing and classifying military and civilian targets after receiving a short command from headquarters — “destroy the naval target approaching the borders of Taiwan”
I apologize in advance if the videos were too brutal for you, I tried to find the least shocking footage. If it is still unacceptable, I will remove the links.
No one except outright tyrants would bomb civilian infrastructure or use nuclear weapons, subjecting millions of people to suffering.
Once serious nuclear weapons are used, everyone dies (to a first approximation), civilian or not. If I recall correctly, it takes about 100 megatons worldwide to casuse nuclear winter and collapse agricultural production.
During the Cold War, the US maintained a position of "strategic ambiguity" on the question of first use. Much of the logic around NATO at the height of the Cold War was based around a first-use nuclear response to overwhelming conventional invasion (see MC 14/13 staged responses). This was the full-scale, Dr Strangelove, batshit-insane "end of civilization" nightmare. Strategic ambiguity was retained around what would trigger each level of response, but the endgame was pretty much total annihilation. I believe France also maintained a separate posture of strategic ambiguity, and they always wanted to ensure a nuclear deterent that didn't rely on NATO.
China and Russia both held official policies of "no first use", but it's uncertain that they would have actually stuck to that in the face of a massively overwhelming conventional invasion.
I want to be clear: The logic of nuclear deterence is just as insane as Dr Strangelove made it out to be. And you may choose to call NATO, the US and France "tyrants"! But they all had policy at least as dangerous as, "Well, we haven't promised that we won't trigger nuclear Armageddon and the death of billions if a large enough number of tanks roll across our borders. Do you feel lucky, punk?"
So as a Westerner, that's a missing piece of the analysis for me. Taiwan has invested heavily in long-range cruise missiles and, in the past, secret nuclear programs. Presumably they had some theory of how they would use that capacity in the face of a massively overwhelming conventional invasion.
And just in case I haven't made it clear, I think MAD is madness. I think even the people who coined the acronym knew that. But when a country is faced with overwhelming conventional invasion, I don't think we can automatically rule it out.
1Am I missing something? It seems like defense surpasses offense in the conventional sense. But if Russia and Ukraine both had nuclear weapons, no drone would be able to prevent that once launched, right? Likewise if I flew an airplane to drop a bunker-buster on a dam (e.g. F-35s are still useful in the Russia/Ukraine war).
Maybe in the limiting case of drone-warfare defense does dominate, but it seems to me that we are some time away from that, in manufacturing capacity alone, if nothing else.
It is well known that tank assaults during World War II were able to break through almost any line of defense, turning warfare from positional into highly maneuverable.
Ukraine has no nuclear weapons; it gave them up in the 1990s at the insistence of the United States (many thanks to the American presidents—they are always on our side). But what would the destruction of a dam or a nuclear bombardment of a city with a million inhabitants give you? It does not allow you to seize territory; instead, you would face international condemnation, the imposition of sanctions, and sometimes a retaliatory nuclear strike, far more powerful.
Ukraine does not even have F-35 Lightning II fighter jets, since these aircraft were not supplied to Ukraine (once again, my deep bow and respect to the American presidents).
Russia is already producing 50,000 FPV drones per month and is capable of doubling that output; this is already comparable to the number of soldiers on the battlefield. One enemy soldier — one drone plus one drone operator, but if AI takes over control on the last mile, it will reduce the burden on drone operators.
https://www.youtube.com/watch?v=tFCbNfGO4Fg
Of course, there are also other longer-range drones (copies of the Iranian Shahed, and China surely has all the blueprints), which are already being stockpiled by the thousands and can overwhelm any air defense system.
https://www.youtube.com/watch?v=5XDiE9UNtiQ
I am trying to describe to you what a war between China and Taiwan would look like — it certainly will not be nuclear and civilian casualties will be minimal (compared to Dresden and Hiroshima), but even that I cannot justify.
Sure, but Ukraine wanted F-35s right, I assume because they thought they would be useful. As to the rest, it seems like you could claim that America 'seized' Japanese territory after a nuclear strike (rewriting the Japanese constitution, occupation, now a staunch ally, etc). Such a strike only has to break the will of the people fighting, or break the ability of command structures to function effectively, you don't have to glass the entire country you want to invade.
In fact, breaching enemy drone defense zones is not impossible:
If military strength is severely imbalanced, one side can suppress enemy drone operators through airstrikes and artillery bombardment;
Armored vehicles equipped with directed-energy weapons, anti-drone weapon stations, and active defense systems can theoretically withstand swarm attacks and penetrate defenses—such as China's Type 100 tank;
Disrupting enemy drone supply chains is a sound strategy. Ukraine's ability to assemble drones using civilian 3D printers stems from its vast strategic depth and imported components from China. These components require complex, large-scale manufacturing facilities—facilities and their logistics chains that are inherently vulnerable.
Future ground warfare will not be entirely dominated by drones: Drone-guided artillery shells, rockets, and aerial bombs will strike hardened targets beyond drone capabilities; Mechanized dog-infantry demolition reconnaissance teams (DRGs) will infiltrate complex terrain to establish incremental area control, with armored units providing direct fire support; across broader fronts, tactical missiles and long-range rockets will hunt self-propelled artillery and destroy supply hubs, while medium-range missiles will neutralize enemy airfields, warehouses, and factories.
Armored vehicles equipped with directed-energy weapons, anti-drone weapon stations, and active defense systems can theoretically withstand swarm attacks and penetrate defenses—such as China's Type 100 tank;
It seems to me that your neural network has over-imagined things. Give it the following task!
Task:
To destroy one M1A1 Abrams tank, only 5 fiber-optic drones were required, which are not susceptible to any interference. How many Abrams tanks should you produce per month if the enemy is producing 50,000 drones?
https://www.youtube.com/watch?v=O2FcqV-M9qM
Your tank-based economy will not withstand such competition, because drones are cheaper.
M1A1 Abrams cost = 10,000,000
Drone cost (each) = 5,000
5 drones = 25,000
Even if China’s Type 100 tank requires 10 drones, it changes nothing. The practice of the war in Ukraine shows that even a drone flight range of 5 km is sufficient to stop any tank column. Tank assaults are a thing of the past — forget them!
I would like to note that this metal “grill” mounted on the tank turret changes its moment of inertia and negatively affects the rotation mechanism, reducing its service life.
In a situation where a drone is cheaper than a soldier, your sabotage and reconnaissance groups (DRGs) will be destroyed just as cheaply as tanks (I have relevant videos, but the horror on the soldier’s face in the final seconds of his life, captured in close-up by a drone camera and transmitted via fiber‑optic cable in Full HD, is excessively shocking even for me, and I will not provide any links), even at night, since thermal imaging cameras are also inexpensive. Although it is possible that a North Korean soldier costs less than $5,000, and that is a truly serious problem — send them humanitarian aid, iPhones, and Netflix series so they can feel the value of their own existence.
Yes, I forgot to mention that Elon Musk's Starlink network, which offers minimal signal latency, would theoretically allow you to operate a VPF drone from a smoothie bar somewhere in Costa Rica. You're going to bomb smoothie bars in Costa Rica—seriously? I'll have another smoothie, please, before the Type 100s go on the attack!
1: As I've repeatedly emphasized across multiple platforms, I did not employ generative AI technology to compose these texts. If they resemble LLM output, it likely stems from my writing style.
2: If tanks can employ directed-energy weapons and cannon-mounted programmed munitions to shoot down hundreds of drones, while striking fortified positions from thousands of meters away under infantry or drone guidance, the enemy assets they destroy and the infantry lives they protect may far outweigh their own cost.
Armor itself serves as an excellent drone deployment platform: it can maneuver upon detection, possesses surplus defensive firepower, and offers at least splinter protection. Without such platforms, drone operators must either remain in rear areas—depleting drone range and reducing sortie frequency—or face certain death upon exposure.
3: DRG units can consist of relatively few humans and numerous robotic platforms, operating covertly whenever possible to minimize drone casualties. If smaller platforms can also deploy effective anti-drone weapons, their casualty rates would be even lower. These teams remain irreplaceable because FPV drones are poorly suited for clearing buildings and tunnels, and struggle to launch attacks from many routes (such as abandoned oil and gas pipelines). Additionally, FPV requires units to mark targets—including by drawing enemy fire—otherwise they prove ineffective against concealed adversaries.
4: Using Starlink to remotely control frontline units is a sound concept but imperfect: During large-scale warfare, frontline units operate in complex electromagnetic environments. You may need to position Starlink receivers tens of kilometers behind the contact line and connect them to frontline units via fiber optics. However, frontline units still require human operators at present.
5: These assumptions are based on cutting-edge technology projected for 2026. Should artificial intelligence advance to solve complex frontline combat challenges, we'll all soon be turned into paperclips.
3These teams remain irreplaceable because FPV drones are poorly suited for clearing buildings...
Drones are perfectly suited for clearing buildings — once again, watch the video:
https://www.youtube.com/watch?v=tFCbNfGO4Fg&t=115s
This is not a person with a camera; it is a fiber-optic FPV drone flying through an open door. The video cuts off at the moment the drone’s warhead detonates.
This once again fully confirms that your response was generated by an LLM that cannot watch or analyze YouTube videos. Try teaching it to do so, so your answer better reflects the actual situation on the battlefield rather than assumptions based on LLM hallucinations.
Back when many expected takeoff in 2027 or so, it was pretty reasonable to assume that the probability of a conflict entirely unrelated to AI was low.[3] But the forecasters behind AI 2027 now expect takeoff in the 2030s.
I'm saddened that this is the takeaway from our new model! It seems misleading. Here is a graph of my+Eli's timelines over time:
https://x.com/DKokotajlo/status/1992316620254155028
1I'm sorry if I misrepresented your beliefs! I was mainly basing this on the timeline in the new model (using Eli's median parameters), in which Automated Coder (AC) is reached in 2031 and ASI in 2034, which I thought was well-described by "takeoff in the 2030s". Does AGI in this plot refer to the same thing as ASI in the model? If your timelines have ASI in 2030 (and presumably AC somewhat earlier), then I can revise what I wrote to try to reflect that.
I guess another issue is that even if the median is in the early 2030s, that still leaves half the probability mass earlier than that, so what I wrote might be misleading if it's taken to imply that you definitely don't expect takeoff before 2030.
I doubt that the potential misconception is actually relevant to the issue of modifying the model due to Taiwan being invaded. Your two forecasts imply that P(automated coder by Jan 2028) is at most 27% and Eli's forecasts have P(automated coder by Jan 2028) is less than 25%. Similarly, neither you nor Eli believe that P(superhuman coders appear by Jan 2029) is at least 40%. Were Taiwan to be invaded during Jan 2027 and extrapolation of pre-invasion trends with exponential growth of compute to yield automated coders after Jan 2029, OpenBrain would no longer have the way to reach the SC level by increasing the compute tenfold. Instead I would expect the USG to merge the companies, have OpenBrain reach Agent-4 far later, try to slow down and realize that they can't gather any more resources. How would DeepCent's compute increase after the invasion?
This point has been made previously here and here, but I think the possibility of war is much more relevant now that people have somewhat longer timelines.
What is especially interesting is the ability of the war to prevent the Slowdown Ending. How likely are the amounts of compute of the two sides to become ~equal, causing both sides to race all the way to doom? Because the AI-2027 forecast by @Daniel Kokotajlo et al would have to be rewritten as follows: Taiwan gets invaded; DeepSeek et al are merged into DeepCent; OpenAI, Anthropic et al unite their efforts into OpenBrain well before[1] misalignment got revealed and either keep racing or fail to regain the lead. DeepCent would also have to race as hard as it can in order to either avoid falling behind or avoid losing its newly found leadership. As a result, mankind either ends up with one side having an aligned AI weaker than the other's misaligned one or[2] with a mutual race.
- ^
The AI-2027 forecast itself had the USA do so in response to Agent-4 being misaligned and the leading company's lobbyists trying to regain the lead.
- ^
Of cource, there also is the alternative of international coordination, but it is far harder due to the invasion.
especially if TSMC's fabs get bombed, but if the conflict isn't centrally about AI I think there's a good chance they'll survive.
I remember reading somewhere that the U.S. / Taiwanese government plan is to render the fabs permanently unusable if Taiwan changes hands. Given the rise in official/unofficial acts of industrial warfare, and the escalating targeting of nonmilitary shipping during the ongoing conflict in Eastern Europe, I would certainly expect this to happen.
Additionally, in regard to industrial sabotage, the historical vulnerability of U.S. infrastructure might be worth considering. Considering the wide array of groups that consider the U.S. government to be their enemy (or at least unworthy of loyalty), not to mention the substantial portion of foreign nationals with loyalty to potentially pro-China home countries, sabotage of U.S. infrastructure could very rapidly hit catastrophic levels if the PRC started offering bounties or leaning on people to act. There have been numerous embarrassing stories of late in which members of the U.S. military who were originally from China sold or gave classified intelligence to China, and there doesn't seem to have been any serious discussion from the security establishment about how to prevent a much more catastrophic version of this from happening during a war.
Depending on your outlook with regard to the competence of the U.S. security establishment, the lack of concern there is indicative of either extreme certainty that a war with China will not occur, or an indicator that such a war might go very badly for the U.S. government much earlier than we would expect. The former means that we might see higher-than-expected stability in the chip market even as the Pax Americana disappears, and the latter means that China would essentially be able to decide the rate of AI progress unilaterally from then on.
I remember reading somewhere that the U.S. / Taiwanese government plan is to render the fabs permanently unusable if Taiwan changes hands.
Yes, to clarify, when I said "I think there's a good chance they'll survive" I was referring to Chinese fabs -- I expect the Taiwanese fabs to be destroyed.
I largely agree with your other points, I think the US is extremely vulnerable to sabotage and has not been taking the issue seriously enough. My main hope in this regard is that China might decide to hold off from targeting civilian infrastructure out of fear of the US's response.
; this discussion is more apocalyptic, predicting global microprocessor production falling to "early 2000s levels for perhaps 15 years
This was from 2022. Since then, the US has made significant efforts in de-risking the semiconductor supply chain. The Arizona fab appears to be ahead of schedule and of course is already operational. Additionally rare earth chokepoints have been identified and begun to be addressed. I would lean towards it being less of a slow down in advanced chip manufacturing than expected in the 2022 report.
1
TL;DR: Most AI forecasts generally assume that if a conflict over Taiwan occurs, it will largely be about AI. I think there's a decent chance for a conflict before either side becomes substantially AGI-pilled.
Thanks to Aaron Scher for comments on a draft of this post.
I'm no China expert, but a lot of China experts seem pretty concerned about the possibility of a conflict over Taiwan. China is currently engaged in a massive military buildup and modernization effort, it's building specialized invasion barges like the Mulberry harbors used in the WWII Normandy landings, and it's conducting amphibious landing exercises with civilian roll-on/roll-off vehicle ferries, many of which China modifies for potential military use. Increasingly frequent military exercises around Taiwan could let China rapidly transition to a full blockade. Its internal propaganda suggests that Taiwanese "provocations" could justify military action, and its leadership continually talk about Taiwan's "return to China", with some even openly discussing "reeducation".
By some cosmic coincidence, 2027, the PLA's centennial, is sometimes identified as the year when the PLA hopes to be ready for a conflict over Taiwan. This doesn't mean China will immediately pull the trigger, but they might want to be prepared by then in case things do escalate. They may believe the next few years represent a window of opportunity before slower growth and a demographic crisis reduce China's power relative to the US. Plus, Xi is 72, and would probably love to cement his legacy by retaking Taiwan in his lifetime.[1]
Manifold currently puts the probability of an invasion of Taiwan by the end of 2027 at around 22%, and before 2030 at around 37%, although I don't think these markets count blockades and other actions that fall short of a full invasion:
Other markets put the chance of a more limited conflict higher:[2]
I'm not trying to make the case here that there will probably be a war. The point I want to make is that while most AI forecasts assume any conflict with China would largely be about AI, I think there's a decent chance a conflict occurs for other reasons before either side becomes AGI-pilled. This point has been made previously here and here, but I think the possibility of war is much more relevant now that people have somewhat longer timelines. Back when many expected takeoff in 2027 or so, it was pretty reasonable to assume that the probability of a conflict entirely unrelated to AI was low.[3] But
the forecasters behind AI 2027 now expect takeoffedit: some of the forecasters behind AI 2027 have median ASI arrival dates in the 2030s.[4] If that's the case, I think there's a good chance Xi decides to escalate over Taiwan before he (or his successor) starts paying serious attention to AI. At the very least, the timelines overlap considerably: developments in AI could shift China's calculus over whether and when to invade, but equally Chinese aggression unrelated to AI could drastically impact AI timelines.It's perfectly reasonable to build a forecasting model that doesn't try to take these kinds of exogenous shocks into account. But I think forecasters should clearly flag when they do this, and ideally provide estimates for how likely they think such events are if the chances are significant. I haven't really seen this in the AI forecasting space: for instance, the AI Futures Project's all-things-considered forecasts don't mention the possibility of a conflict, and this scenario has China blockading Taiwan as late as 2034, and only in response to US cyberattacks on Chinese AI development.
I also think the chances of conflict are high enough that it would be very valuable to have forecasts specifically focused on understanding AI timelines in the event of a war. There's been some discussion of this here and here, but those are
ancient historyover two years old at this point, so something more up-to-date would be useful. I'll give a few of my thoughts below, but this is mostly just speculation -- I'd really like to see modeling work by more knowledgeable people on how a conflict would impact AI.Impacts on compute
In general, a conflict over Taiwan would almost certainly slow down progress in AI by disrupting access to compute. This report estimates 20 months for other firms to catch up to TSMC; this discussion is more apocalyptic, predicting global microprocessor production falling to "early 2000s levels for perhaps 15 years." It's less clear to me who would come out ahead in relative terms, though. The US is very reliant on Taiwan, but I think in some cases it might actually be able to maintain a good portion of its compute advantage even if Taiwanese production is cut off, because China also depends on Taiwan for much of its compute. I'm pretty uncertain about this, though.
If things escalate to a shooting war, it's likely that much of Taiwan's manufacturing will be destroyed by one side or the other. If China manages to take over Taiwan, they might gain access to some of TSMC's expertise or technology even if the fabs are destroyed. A good part of TSMC's edge is reportedly in its engineering talent and in knowledge held by a small number of individuals. But I think China would probably have a lot of difficulty gaining the cooperation of TSMC employees or recreating the talent base after a war, so I don't expect China to benefit very much: the main effect would be everyone losing access to Taiwanese production.
Meanwhile, I don't think Chinese chip production would be disrupted as much by a war. A US blockade or other wartime supply-chain disruptions would certainly make things more difficult, but it seems much harder to cut off Chinese production without just bombing the fabs. The US might do that,[5] especially if TSMC's fabs get bombed, but if the conflict isn't centrally about AI I think there's a good chance they'll survive.
But Chinese chip production currently isn't very competitive and might not catch up for some time. If China is still reliant on foreign chips, the conflict could see the US maintain its compute advantage because it would almost certainly stop selling China chips and crack down on chip smuggling. While the US would likely lose access to Taiwan's production, at least for the duration of the conflict, it would still have TSMC's Arizona fab plus whatever production Intel can manage.[6] I think the relative compute balance here really depends on how quickly US domestic production ramps up compared to Chinese production -- modeling this properly would be very valuable!
A major risk, though, is that Taiwanese production might be used as a bargaining chip if it's not destroyed. This could be as part of a peace deal after a war, but it could also happen before a full-scale war starts. China might impose a blockade, take some outlying islands, or use other kinds of pressure to try to force a capitulation or extract concessions without having to invade. It seems unlikely that China would be able to take over completely without a fight, but it could gain better access to Taiwanese production: the US might agree to loosen export controls or even give China access to TSMC or ASML's tech as part of a settlement. If the US government isn't AGI-pilled at this point, it might not even value this particularly highly, or view it as opening up a market for American companies.
Securitization
One other scenario worth considering is that US or Chinese leaders might start to wake up to AGI during the conflict. An intelligence explosion probably increases the risk of war in the best of times; if there's already a war going on, things could get very ugly. If AI gets securitized we'd likely see attacks on fabs and data centers, secret Manhattan Projects, assassinations, and little room for safety research, let alone any sort of coordination on AI. On the other hand, if the US and China sabotage each other hard enough this could end up delaying AGI significantly.[7]
Conclusion
Of course China might decide not to invade, or we might get AGI first, rendering all of this moot. But I think the chance of a conflict over Taiwan years before AGI is high enough that it should probably be factored into peoples' timelines. It's easy to forget, but people care about other things besides AGI! And the decisions they make could have big impacts on the AI race; we've seen this repeatedly with chip sales to China, and we could very well see it again.
And despite recent talk, I don't think he genuinely expects to live to 150, although I suppose it would explain his apparent lack of succession planning.
I don't think you should pay much attention to my probabilities because I'm neither a China expert nor an experienced forecaster, but for the record, I think 22% for invasion by the end of 2027 is maybe a bit high (and I've bet on this), while the other markets roughly match my estimates.
In the original AI 2027 timeline the CCP contemplates an invasion or blockade of Taiwan, but only in response to the US advantage in compute.
As pointed out in the comments, this is more true for some forecasters than for others, and their models still put substantial probability on ASI being developed before 2030. If you're pretty confident that we'll be well on our way to ASI before 2030, you should probably put less stock on the arguments in this post. My impression, though, is that many people do think there's a substantial chance of takeoff in the 2030s, in which case I think it's worth considering how a conflict would impact timelines.
Can the US bomb the Chinese mainland without triggering a nuclear war? I don't know! China maintains a no first use policy, but who knows how that will hold up during a war.
China could try to disable the US fabs (e.g. with cyberattacks), but the US would likely retaliate against Chinese fabs, at which point I'm not sure anyone's left making chips. I guess in that case the US's larger preexisting stock of compute might give it an advantage.
I'm not sure I'd want to stay in the Bay Area for this, though -- I don't want to get "sabotaged."