https://www.youtube.com/watch?v=44S4C6axNXg
|
Transcript 0:02 [Music] 0:08 welcome to fyi the four-year innovation 0:10 podcast 0:11 this show offers an intellectual 0:13 discussion on technologically enabled 0:15 disruption because investing in 0:17 innovation starts with understanding it 0:19 to learn more visit arc dash invest dot 0:22 com 0:23 [Music] 0:26 arc invest is a registered investment 0:28 advisor focused on investing in 0:30 disruptive innovation this podcast is 0:32 for informational purposes only and 0:33 should not be relied upon as a basis for 0:35 investment decisions it does not 0:37 constitute either explicitly or 0:38 implicitly any provision of services or 0:40 products by arc all statements made 0:42 regarding companies so securities are 0:44 strictly beliefs and points of view held 0:45 by arc or podcast guests and are not 0:48 endorsements or recommendations by arc 0:50 to buy sell or hold any security clients 0:52 of arc investment management may 0:54 maintain positions in the securities 0:55 discussed in this podcast 0:57 [Music] 0:59 so we're so excited that we just 1:00 published our latest tesla valuation 1:03 model for our 2026 price target um i 1:06 have here with me i'm tasha keeney i'm 1:09 here with sam korres and will summerlin 1:12 we're all arc analysts tesla is a great 1:15 example of how our research converges at 1:18 arc so will is our ai specialist of 1:22 course you know sam an expert on 1:25 electric vehicles and their batteries 1:27 and i cover autonomous technology and 1:29 autonomous driving 1:31 so uh first we'll kick it off by 1:33 explaining some of the the background of 1:36 our top down research and and how it 1:38 applies here and then we'll get into uh 1:40 the model itself 1:42 sounds good and you know just tying into 1:45 ark's research 1:47 ecosystem here in the convergence right 1:49 we have top-down models for electric 1:51 vehicles we've 1:53 got will he's talking about ai and 1:56 foundation models and training there and 1:59 then tasha from autonomous so 2:01 we all get together and 2:03 brainstorm and that's that's how we make 2:05 the tesla model which we'll get to but i 2:08 guess to kick things off we'll start on 2:09 the electric vehicle side of things 2:12 and our top line forecast for electric 2:15 vehicles 2:16 is that sales will increase roughly 2:18 eight-fold from 4.8 million units in 2:21 2021 to 40 million units in 2026 and 2:26 that's a global number 2:28 and that'd be roughly 53 compound annual 2:31 growth rate each year and what's really 2:34 driving this is battery cost declines 2:37 and so 2:38 we use wright's law and this is across 2:41 many different themes and technologies 2:44 but for batteries it's so important 2:46 because the battery is the single 2:49 largest cost component of the vehicle 2:51 and so wright's law states that for 2:53 every cumulative doubling of production 2:55 you get a fixed percent cost decline and 2:58 for 2:58 lithium-ion batteries that's roughly 28 3:02 and so we think that by 2023 a 3:06 electric vehicle is going to reach 3:08 sticker price parity so that's the 3:10 upfront cost the consumer sees with a 3:12 like for like gas-powered car and at 3:14 that point 3:15 it should really be a no-brainer and we 3:18 think that we're going to see a huge 3:20 shift in demand towards electric 3:22 vehicles 3:24 with that i'll pass it to will to talk 3:26 about 3:26 ai 3:28 so one of the opportunities we're really 3:29 excited about within ai is the 3:31 opportunity for foundation models and to 3:34 dig into this a little bit one of the 3:36 things that we've seen historically is 3:38 that the larger models are ai models the 3:40 better they perform 3:42 and this is a trend we think we'll 3:43 continue we've seen this with models 3:45 like gpt3 which is trained on most of 3:48 the text from the internet um we've seen 3:50 this across dali and bart and a number 3:52 of other large language models we think 3:54 that the same principle will apply uh 3:57 within the other application of ai which 3:59 is helping robots move through the 4:00 physical world 4:02 and tesla has really unique data assets 4:04 in that they've seen billions of miles 4:06 across their fleet and use that to train 4:08 their own models for full self-driving 4:09 and autonomy we think there's actually 4:11 an opportunity for tesla to make those 4:13 models openly available so that other 4:16 developers other companies can actually 4:18 build applications and products on top 4:20 of tesla's models and this is what we 4:22 call a foundation model 4:24 so based on our research and big ideas 4:26 we think the opportunity for foundation 4:28 models um could be worth about 25 4:30 trillion dollars in terms of enterprise 4:32 value by 2030 and we think the market 4:34 will bifurcate there will be models that 4:36 are used to solve problems in sort of 4:39 digital native contacts and so these are 4:41 models like gpt3 and dali that can help 4:44 edit text or manipulate images uh or 4:47 generate stories and then we think 4:49 there'll be a different use case which 4:51 is helping robots move through the 4:52 physical world and we think that'll be a 4:53 different set of foundation models so 4:56 while tesla hasn't actually openly 4:58 discussed this strategy we do think it's 5:00 a possibility especially given elon's 5:02 excitement around ai and given their 5:04 unique data assets especially given 5:06 their investments in compute 5:08 and so we're very excited to see how 5:09 that opportunity could manifest 5:11 with that turn it back to sam and tasha 5:13 thanks well 5:15 and you know one of the the biggest 5:16 potential markets that we see in ai is 5:19 autonomous driving and we think that all 5:23 vehicles in the future all passenger 5:25 cars will be electric so when we talk 5:27 about autonomous cars they're electric 5:29 autonomous cars 5:30 ultimately why we're so interested in 5:32 autonomous driving is because we think 5:34 that it could dramatically lower the 5:36 cost to get around 5:38 so an autonomous taxi could price as low 5:42 as 25 cents per mile that's less than 5:45 half of the cost of driving a personal 5:47 car in the us a new personal car that's 5:50 really dramatic it's going to bring a 5:52 bunch of people into the ride hill 5:53 market that are not customers today 5:56 because of that extremely low cost point 5:58 of course cars will be a lot safer too 6:00 but ultimately we think cost is going to 6:01 drive demand 6:03 and we think that this is such a large 6:05 opportunity 6:06 on a revenue basis the total addressable 6:09 market is about 11 trillion dollars and 6:12 and that's really our updated work on 6:14 autonomy 6:15 so it could be as cheap as 25 cents per 6:18 mile for a ride but we also see a lot of 6:20 demand at the you know 50 cents to a 6:23 dollar range um for uh existing uh 6:27 customers today that are either you know 6:29 traveling around for their own personal 6:31 needs or for work related travel we've 6:34 done some analysis on how customers 6:35 value their time uh to dimension those 6:38 price points 6:39 ultimately we think that uh the the 6:41 platform providers or the companies that 6:44 own the technology stack that allow cars 6:46 to be autonomous could enjoy enjoy an 6:50 enterprise value 6:51 of 11 to 12 trillion dollars by 2026 so 6:55 this is really a massive opportunity and 6:57 one that tesla is uniquely uh well set 7:00 up to to participate in as it's both a 7:02 vehicle manufacturer and potentially an 7:05 autonomous platform provider and with 7:07 that uh let's get into the model yeah 7:10 tasha let's let's just get to the to the 7:12 meat of it right off the bat what is 7:14 arcs 2026 price target for tesla and 7:18 then we'll go into the methodology right 7:20 so our expected value uh for tesla stock 7:24 price in 2026 is 7:27 4600 per share 7:29 we got there by 7:31 we have a monte carlo simulation that 7:34 we've run a million times and the 7:37 expected value is the average of of all 7:39 of those simulations 7:41 and then how does arc get to its 7:45 upside and downside cases 7:48 our bear case for tesla is uh 2 900 per 7:52 share in 2026 and that is the 25th 7:56 percentile of that million simulation 7:59 run of a mrt carlo analysis so in other 8:02 words we think 8:04 that 8:04 25 percent of the time 8:07 tesla could be worth 2 900 per share or 8:11 less and then our bull price per share 8:14 is 5 8:15 800 8:16 by 2026. so that is the 75th percentile 8:20 or in other words 8:22 we think that there's a 25 probability 8:25 that tesla could be worth five thousand 8:27 eight hundred hundred dollars per share 8:28 or more uh by 2026. so ultimately you 8:32 know this is all again a result of the 8:34 monte carlo analysis and our monte carlo 8:36 model is built up of 38 input variables 8:40 uh which we give a range of 8:42 possibilities within the model 8:45 and ultimately these are the factors 8:46 that we think could have some impact on 8:49 tesla's five-year price target 8:52 you know 38 inputs is a lot here uh 8:55 i think you know when i when we built 8:57 the model we then you know tested and we 9:00 found the the key inputs so so what are 9:02 those five key inputs that 9:05 uh we think are really the main drivers 9:07 of this model 9:09 so our five key inputs are capital 9:12 efficiency 9:13 how the the maximum production increase 9:17 that tesla can get in any given year so 9:19 how much they're allowed they're they're 9:21 able to increase production year over 9:23 year on a percentage basis 9:26 we also have an input for uh how many 9:28 cars we think will be on the the ride 9:31 hill network uh by 2026 this is um the 9:34 potential for tesla to launch a 9:35 human-driven ride hill network um by 9:38 that point um and then we also have the 9:42 uh estimated launch year for robo taxis 9:45 so for autonomous ride hail when do we 9:47 think that launch could potentially 9:49 happen 9:50 so while we do you know those are our 9:52 five key inputs out of the 38 and then 9:55 you know you'll notice as you play 9:57 around with the model which is open 9:58 sourced and available on github that the 10:00 ones that really affect it the most are 10:03 our assumptions around capital 10:04 efficiency and autonomy 10:07 great so before we dive into autonomy 10:09 because that's that's really where the 10:11 bulk of the value is coming from maybe 10:14 i'll just touch on 10:16 capital efficiency and how we're looking 10:18 at it 10:19 and some context around that 10:21 and so originally when we were doing 10:23 research on electric vehicles you know 10:26 we're saying what is the capital 10:27 efficiency or how much money needs to be 10:29 spent for an incremental unit of 10:32 capacity 10:34 to be built in the 10:37 auto industry as it stands and that 10:38 number was fourteen thousand dollars 10:41 and so that's kind of what we had used 10:43 and said okay this is the baseline if 10:45 tesla 10:46 matches industry average and so if you 10:50 go and you look in the model 10:52 and our blog you can see that when tesla 10:55 first started ramping the model 3 10:58 their capital efficiency was far far 11:00 worse they were at roughly eighty four 11:02 thousand dollars per incremental unit of 11:04 capacity but they 11:07 dramatically improved and in 2021 they 11:10 were down to seven thousand seven 11:12 hundred dollars per incremental unit of 11:13 capacity which is pretty 11:16 remarkable right already half of what 11:18 the industry average was 11:20 and so in the monte carlo 11:23 we say that in the downside case 11:25 they just marginally improve and and 11:28 stay at roughly seven thousand dollars 11:29 per unit incremental unit capacity 11:32 but in our upside case we have that cost 11:35 coming down to two thousand dollars and 11:38 i think what's what's really interesting 11:39 is if you play with the model you can 11:41 see that you know this input itself 11:44 is actually not as meaningful as it was 11:46 in the past 11:48 because 11:49 tesla is no longer capital constrained 11:52 and so what that means is that you know 11:54 their ability to scale production is not 11:58 based on how much money they have in the 12:00 bank account how much money they can 12:01 spend it's really based on 12:04 management bandwidth whether or not 12:05 there's supply constraints which have 12:08 been particularly relevant over the past 12:10 year 12:11 and so that's where this other key input 12:13 comes into play 12:15 which is the maximum annual production 12:17 increase 12:18 and you know this is 12:20 probably one of the bigger drivers on 12:23 the vehicle production side of it 12:26 but then you know tasha 12:28 i'll hand it to you to talk about 12:30 autonomous because really the the launch 12:32 here and when adoption happens i think 12:35 is 12:36 is the biggest single biggest driver for 12:38 this model 12:39 thanks sam robo taxis do have you know a 12:42 very large impact 12:44 on our valuation target um you'll notice 12:47 that in the blog we've broken out that 12:49 they will contribute approximately 60 12:53 of the expected value in the model um 12:56 can be attributed to 12:57 robo taxis um and the we've changed a 13:01 little bit the way that we've we've 13:02 modeled this versus last year um so last 13:04 year we had a single input which was the 13:07 probability of robo-taxi launch within 13:09 the five-year window 13:12 but we changed that this year to have we 13:14 still have that probability input um but 13:16 we also have the robo-taxi launch year 13:19 input 13:20 and effectively this is just because i 13:23 think this is a bit a little bit closer 13:24 to how people think of you know whether 13:26 or not tesla will launch robo taxis it's 13:29 okay if it happens which year will it 13:30 happen if at all 13:32 and so you'll notice that uh you know in 13:34 our downside case uh we're assuming that 13:36 tesla launches in 2026 13:39 so that would be four years after elon 13:41 musk has said that they he expects to 13:44 solve for full autonomy and in the the 13:46 upside case we're assuming a launch year 13:49 of 2023 13:50 so again that would be actually one year 13:52 after um what elon has said could 13:54 possibly happen 13:56 and the mid at the midpoint of our 13:58 assumptions it's the 2024 year so about 14:02 two years later than expected by the 14:04 company 14:05 and i also just want to point out that 14:07 nine percent of the time in the model uh 14:10 the launch year actually falls out of 14:13 the window that we're forecasting out of 14:15 that five years so effectively in nine 14:18 percent of our simulations were 14:20 including no revenue um for autonomous 14:23 taxis at all 14:25 another thing that we updated uh that i 14:27 alluded to earlier is how we think of 14:30 the price points that we'll see in the 14:32 autonomous ride hill market so you know 14:35 when we first did our analysis on 14:37 autonomous taxis we saw that they could 14:38 be priced as low as 25 cents per mile 14:41 but ultimately we think that um you know 14:43 we've seen bride hill has proven that 14:45 there is demand around the two dollars 14:47 to four dollars per mile price range 14:50 that we see today 14:51 we think that um because consumers 14:55 in western markets our analysis shows 14:57 that they value their time at roughly 14:59 equivalent to 15:01 average hourly wage for 15:03 miles driven uh with the purpose of 15:05 going to work so for commuting miles 15:08 at roughly equivalent to hourly wage and 15:10 then for non-commuting miles or driving 15:13 that you're doing for your personal use 15:14 that's valued at roughly half of hourly 15:16 wage so that gives a lot a lot of demand 15:20 in western markets at price points of 15:23 about a dollar 10 to 60 cents 15:26 and then in 15:28 lower income countries we expect 15:30 significant demand at the 50 cent mile 15:33 mark um that's because if you look at uh 15:36 you know how deedy is already pricing 15:38 ride hail options in in places like 15:40 china today it's already pretty 15:41 inexpensive at about 50 cents so 15:44 ultimately the 25 cents per mile price 15:46 point comes in 15:47 and we think that this this will be the 15:49 long tail of of additional customers 15:52 that will now be brought into the ride 15:54 hill market that were not there before 15:57 that's where um you know we see a 15:58 dramatic change 16:00 in habits and again in additional people 16:02 coming into the market for this 16:04 extremely cheap service so ultimately 16:07 you know all of those price points get 16:08 factored into our model uh we're 16:10 assuming that once tesla penetrates a 16:13 certain segment of the market uh if they 16:15 keep on adding autonomous miles they 16:17 fall into the next uh price point 16:19 segment and that's ultimately how you 16:20 can think of it factoring in 16:22 um so 16:24 this 16:25 what's the end effect well we think that 16:27 there's more opportunity or more revenue 16:29 opportunity in autonomous driving than 16:31 we previously thought there was 16:33 so tasha as people go through the model 16:36 and play with it 16:37 um one it's like if you're gonna do that 16:39 you know download this on github uh but 16:42 if people are gonna do that you know and 16:44 they have questions or 16:46 uh they don't understand something 16:47 what's the best way for people to to 16:50 reach out and to interact with us 16:51 because you know a big part of ark's 16:53 research is this open sourcing 16:56 and research ecosystem so what's the 16:58 best way for people to to get 17:01 involved and interact 17:03 good question there's really two ways so 17:05 for our github nerds out there you can 17:08 definitely post on github 17:11 use the the comments while there to to 17:14 raise any issues that you see with the 17:16 model or the problem section 17:18 and also tweet at us you know i am tasha 17:22 arc 17:22 uh sam your your 17:25 handle is 17:27 s chorus arc 17:29 and our director of research b bewinton 17:31 arc 17:33 we are all available to answer questions 17:35 on the model and we'd love to hear your 17:37 feedback 17:39 arc believes that the information 17:41 presented is accurate and was obtained 17:42 from sources that arc believes to be 17:44 reliable however arc does not guarantee 17:46 the accuracy or completeness of any 17:48 information and such information may be 17:50 subject to change without notice from 17:51 arc historical results are not 17:53 indications of future results certain of 17:55 the statements contained in this podcast 17:57 may be statements of future expectations 17:59 and other forward-looking statements 18:00 that are based on arc's current views 18:02 and assumptions and involve known 18:03 unknown risks and uncertainties that 18:05 could cause actual results performance 18:07 or events to differ materially from 18:08 those expressed or implied in such 18:10 statements 18:15 [Music] 18:16 you
|
0:02 [音樂] 0:08 歡迎來到四年創新 0:10 播客 0:11 這個節目提供了一個知識分子 0:13 技術支持的討論 0:15 中斷，因為投資 0:17 創新始於理解它 0:19 要了解更多信息，請訪問 arc dash 投資點 0:22 com 0:23 [音樂] 0:26 arcinvest是註冊投資 0:28 專注於投資的顧問 0:30 這個播客是顛覆性的創新 0:32 僅供參考和 0:33 不應以此為依據 0:35 投資決策它不 0:37 構成明確或 0:38 隱含地提供任何服務或 0:40 產品 by arc 所有聲明 0:42 關於公司，因此證券是 0:44 嚴格持有的信念和觀點 0:45 通過弧或播客嘉賓，而不是 0:48 arc的認可或推薦 0:50 買入賣出或持有任何證券客戶 0:52 弧投資管理可能 0:54 維持證券頭寸 0:55 在這個播客中討論過 0:57 [音樂] 0:59 所以我們很興奮，我們只是 1:00 發布了我們最新的特斯拉估值 1:03 我們 2026 年價格目標的模型 1:06 和我在一起，我是塔莎基尼，我是 1:09 和 sam korres 和 will summerlin 在這裡 1:12 我們都是電弧分析師 特斯拉很棒 1:15 我們的研究如何融合的例子 1:18 arc so will 是我們的人工智能專家 1:22 當然你知道山姆是一位專家 1:25 電動汽車及其電池 1:27 我涵蓋自主技術和 1:29 自動駕駛 1:31 所以首先我們會開始 1:33 解釋一些背景 1:36 我們自上而下的研究以及它是如何進行的 1:38 適用於此，然後我們將進入呃 1:40 模型本身 1:42 聽起來不錯，你知道只是配合 1:45 方舟的研究 1:47 這裡的生態系統在匯聚權 1:49 我們有自上而下的電動模型 1:51 我們擁有的車輛 1:53 他會不會在談論人工智能和 1:56 基礎模型和培訓 1:59 然後來自自治的tasha 2:01 我們都聚在一起 2:03 頭腦風暴，這就是我們製作的方式 2:05 我們將使用的特斯拉模型，但我 2:08 猜想開始我們將開始 2:09 電動汽車的一面 2:12 以及我們對電動汽車的收入預測 2:15 汽車 2:16 是銷售額會大致增加嗎 2:18 是 480 萬台的 8 倍 2:21 2021 年到 2026 年達到 4000 萬台， 2:26 這是一個全球數字 2:28 這將是大約 53 複合年 2:31 每年的增長率以及實際情況 2:34 推動這是電池成本下降 2:37 所以 2:38 我們使用賴特定律，這是跨越 2:41 許多不同的主題和技術 2:44 但對於電池來說，它是如此重要 2:46 因為電池是單體 2:49 車輛的最大成本組成部分 2:51 所以賴特定律指出，對於 2:53 產量每累計翻一番 2:55 你得到一個固定百分比的成本下降和 2:58 為了 2:58 鋰離子電池大約 28 3:02 所以我們認為到 2023 年 3:06 電動汽車將達到 3:08 標價平價所以這就是 3:10 消費者看到的前期成本 3:12 喜歡汽油動力汽車和 3:14 那一點 3:15 這真的應該是不費吹灰之力的，我們 3:18 認為我們會看到一個巨大的 3:20 需求向電動化轉變 3:22 汽車 3:24 我會把它傳給will to talk 3:26 關於 3:26 艾 3:28 所以我們真正的機會之一 3:29 興奮內ai是 3:31 基礎模型的機會，並 3:34 深入研究其中之一 3:36 我們在歷史上看到的事情是 3:38 較大的模型是ai模型 3:40 他們表現得更好 3:42 這是我們認為我們會的趨勢 3:43 繼續我們已經在模型中看到了這一點 3:45 像 gpt3 一樣，它受過大部分訓練 3:48 來自互聯網的文字我們已經看到了 3:50 這橫跨大理和巴特和一個數字 3:52 我們認為的其他大型語言模型 3:54 同樣的原則也適用 3:57 在人工智能的另一個應用程序中 3:59 正在幫助機器人通過 4:00 物理世界 4:02 特斯拉擁有真正獨特的數據資產 4:04 因為他們已經看到了數十億英里 4:06 在他們的艦隊中並用它來訓練 4:08 他們自己的全自動駕駛模型 4:09 我們認為實際上有自主權 4:11 特斯拉有機會製造這些 4:13 模型公開可用，以便其他 4:16 其他公司的開發人員實際上可以 4:18 在上面構建應用程序和產品 4:20 特斯拉的模型，這就是我們 4:22 調用基礎模型 4:24 所以基於我們的研究和大創意 4:26 我們認為基礎的機會 4:28 模型 um 可能值 25 左右 4:30 萬億美元的企業 4:32 到 2030 年的價值，我們認為市場 4:34 會分叉 會有模型 4:36 用於解決某種問題 4:39 數字原生聯繫人，所以這些是 4:41 可以提供幫助的 gpt3 和 dali 等模型 4:44 編輯文本或操作圖像呃或 4:47 產生故事然後我們思考 4:49 會有一個不同的用例 4:51 正在幫助機器人通過 4:52 物理世界，我們認為這將是一個 4:53 不同的基礎模型集 4:56 雖然特斯拉實際上並沒有公開 4:58 討論了這個策略，我們確實認為它是 5:00 尤其是考慮到埃隆的可能性 5:02 圍繞 ai 的興奮和考慮到他們的 5:04 獨特的數據資產，特別是給定的 5:06 他們在計算方面的投資 5:08 所以我們很高興看到如何 5:09 這個機會可能會顯現 5:11 把它轉回山姆和塔莎 5:13 非常感謝 5:15 你知道最大的之一 5:16 我們在人工智能中看到的潛在市場是 5:19 自動駕駛，我們認為所有 5:23 未來的車輛所有乘客 5:25 汽車將是電動的，所以當我們交談時 5:27 關於自動駕駛汽車，它們是電動的 5:29 自動駕駛汽車 5:30 最終為什麼我們如此感興趣 5:32 自動駕駛是因為我們認為 5:34 它可以顯著降低 5:36 出行費用 5:38 因此自動駕駛出租車的價格可能會低至 5:42 每英里 25 美分，低於 5:45 個人駕駛費用的一半 5:47 car in the us 一輛新的私人汽車 5:50 真的很戲劇化，它會帶來一個 5:52 一群人進入騎山 5:53 今天不是客戶的市場 5:56 因為那個極低的成本點 5:58 當然汽車也會更安全 6:00 但最終我們認為成本會 6:01 推動需求 6:03 我們認為這是一個很大的 6:05 機會 6:06 以收入為基礎的總可尋址 6:09 市場規模約為 11 萬億美元， 6:12 這真的是我們更新的工作 6:14 自治 6:15 所以它可能便宜到每人 25 美分 6:18 一英里，但我們也看到了很多 6:20 你知道 50 美分到一個 6:23 美元範圍 um for uh 現有 uh 6:27 今天的客戶要么是你認識的 6:29 為自己的私人旅行 6:31 需要或與工作相關的旅行，我們已經 6:34 對客戶如何做一些分析 6:35 珍惜他們的時間 uh 來衡量那些 6:38 價格點 6:39 最終我們認為呃 6:41 平台提供商或公司 6:44 擁有允許汽車的技術堆棧 6:46 自主可以享受 6:50 企業價值 6:51 到 2026 年將達到 11 到 12 萬億美元 6:55 這確實是一個巨大的機會 6:57 一輛特斯拉獨一無二的 7:00 最多參加，因為它既是一個 7:02 汽車製造商和潛在的 7:05 自治平台提供商並與 7:07 嗯，讓我們進入模型是的 7:10 tasha 讓我們開始吧 7:12 它的肉是什麼 7:14 特斯拉的 arcs 2026 價格目標和 7:18 然後我們將進入正確的方法 7:20 所以我們對特斯拉股票的期望值 7:24 2026年的價格是 7:27 每股4600 7:29 我們到了那裡 7:31 我們有一個蒙特卡羅模擬 7:34 我們已經運行了一百萬次並且 7:37 期望值是所有的平均值 7:39 那些模擬 7:41 然後弧是如何到達它的 7:45 上行和下行案例 7:48 我們的特斯拉熊箱是 uh 2 900 per 7:52 分享到 2026 年，也就是第 25 次 7:56 百萬模擬的百分位數 7:59 在其他中運行mrt carlo分析 8:02 我們認為的話 8:04 那 8:04 25% 的時間 8:07 特斯拉可能價值 2 900 每股或 8:11 更少，然後是我們的每股牛市價格 8:14 是 5 8:15 800 8:16 到 2026 年。所以這是第 75 個百分位 8:20 或者換句話說 8:22 我們認為有 25 的概率 8:25 特斯拉可能值五千 8:27 每股八百美元 8:28 到 2026 年甚至更多。所以最終你 8:32 知道這一切又是 8:34 蒙特卡羅分析和我們的蒙特卡羅 8:36 模型由 38 個輸入變量構成 8:40 呃，我們給出了一個範圍 8:42 模型中的可能性 8:45 最終這些是因素 8:46 我們認為可能會對 8:49 特斯拉的五年目標價 8:52 你知道 38 個輸入在這裡很多 8:55 我想你知道我們什麼時候建造的 8:57 我們然後你知道的模型測試了，我們 9:00 找到了關鍵輸入所以是什麼 9:02 這五個關鍵輸入 9:05 呃，我們認為真的是主要驅動力 9:07 這個模型的 9:09 所以我們的五個關鍵投入是資本 9:12 效率 9:13 最大產量如何增加 9:17 特斯拉可以在任何一年獲得，所以 9:19 他們被允許多少 9:21 能夠增加一年的產量 9:23 年百分比 9:26 我們也有一個輸入，呃多少 9:28 我們認為會在路上行駛的汽車 9:31 希爾網絡 呃，到 2026 年，這是 9:34 特斯拉有可能推出一款 9:35 人類駕駛的騎山網絡 um by 9:38 那一點，然後我們也有 9:42 呃，機器人出租車的預計發布年份 9:45 所以對於自動駕駛冰雹我們什麼時候 9:47 認為發射可能 9:49 發生 9:50 所以當我們知道這些是我們的 9:52 從 38 個按鍵中輸入 5 個按鍵，然後 9:55 你知道你會在玩的時候注意到 9:57 與打開的模型一起 9:58 來源並在 github 上可用 10:00 真正影響它最大的是 10:03 我們對資本的假設 10:04 效率和自主權 10:07 太好了，所以在我們深入自主之前 10:09 因為那才是真正的地方 10:11 大部分價值可能來自 10:14 我會談談 10:16 資本效率和我們的看法 10:18 在它 10:19 以及周圍的一些背景 10:21 所以最初我們在做的時候 10:23 你知道的電動汽車研究 10:26 我們說什麼是首都 10:27 效率或需要多少錢 10:29 花費了一個增量單位 10:32 容量 10:34 建在 10:37 汽車工業的現狀 10:38 數字是一萬四千美元 10:41 這就是我們使用的那種 10:43 並說好吧，如果這是基線 10:45 特斯拉 10:46 符合行業平均水平，因此如果您 10:50 去看看模型 10:52 和我們的博客你可以看到當特斯拉 10:55 首先開始加速模型 3 10:58 他們的資本效率遠遠不夠 11:00 更糟糕的是，他們大約八十四歲 11:02 千元/增量單位 11:04 能力，但他們 11:07 顯著改善，並在 2021 年 11:10 下降到七千七 11:12 百元/增量單位 11:13 可愛的容量 11:16 了不起的權利已經一半 11:18 行業平均水平是 11:20 所以在蒙特卡洛 11:23 我們說在不利的情況下 11:25 他們只是略微改善，並且 11:28 停留在大約七千美元 11:29 每單位增量單位容量 11:32 但在我們的上行情況下，我們有這個成本 11:35 降到兩千美元 11:38 我認為真正有趣的是什麼 11:39 如果你和模型一起玩，你可以 11:41 看到你知道這個輸入本身 11:44 實際上沒有以前那麼有意義 11:46 在過去 11:48 因為 11:49 特斯拉不再受資金限制 11:52 所以這意味著你知道 11:54 他們的規模生產能力不 11:58 根據他們有多少錢 12:00 銀行賬戶可以存多少錢 12:01 花費它真的是基於 12:04 是否管理帶寬 12:05 存在供應限制 12:08 在過去特別重要 12:10 年 12:11 這就是其他關鍵輸入的地方 12:13 發揮作用 12:15 這是最大年產量 12:17 增加 12:18 你知道這是 12:20 可能是更大的驅動因素之一 12:23 它的車輛生產方面 12:26 但是你知道塔莎 12:28 我把它交給你談談 12:30 自主，因為真的是發射 12:32 我想在這里和收養髮生時 12:35 是 12:36 是最大的單一最大驅動力 12:38 這個模型 12:39 謝謝山姆機器人出租車你知道嗎 12:42 影響很大 12:44 在我們的估值目標上，你會注意到 12:47 在博客中我們已經打破了 12:49 他們將貢獻大約 60 12:53 模型中的期望值 um 12:56 可以歸因於 12:57 機器人出租車嗯，我們已經改變了 13:01 有點像我們的方式 13:02 與去年相比，這個模型是最後一個 13:04 一年我們有一個單一的輸入，即 13:07 機器人出租車發射的概率 13:09 五年窗口 13:12 但今年我們改變了這一點 13:14 仍然有那個概率輸入，但是 13:16 我們還有機器人出租車發布年 13:19 輸入 13:20 實際上這只是因為我 13:23 認為這有點接近 13:24 人們對你的看法知道是否 13:26 特斯拉是否會推出自動駕駛出租車 13:29 好吧，如果它發生在哪一年 13:30 如果有的話 13:32 所以你會注意到你知道 13:34 我們的不利情況呃我們假設 13:36 特斯拉將於 2026 年推出 13:39 所以那將是埃隆之後的四年 13:41 馬斯克曾表示，他們希望 13:44 解決完全自治並在 13:46 上行情況我們假設發布年份 13:49 2023 年的 13:50 所以那實際上是一年 13:52 嗯，埃隆說過的話可以 13:54 可能發生 13:56 和我們中點的中點 13:58 假設這是 2024 年 14:02 比預期晚兩年 14:04 公司 14:05 我也只想指出 14:07 百分之九的時間在模型中 14:10 發射年份實際上不屬於 14:13 我們預測的窗口 14:15 這五年如此有效地在九 14:18 我們模擬的百分比是 14:20 包括自治的沒有收入 14:23 出租車 14:25 我們更新的另一件事，呃，我 14:27 前面提到的是我們是怎麼想的 14:30 我們將在 14:32 自動駕駛山市場，所以你知道 14:35 當我們第一次對 14:37 我們看到的自動出租車，他們可以 14:38 價格低至每英里 25 美分 14:41 但最終我們認為你知道 14:43 我們已經看到新娘山已經證明了 14:45 兩美元左右有需求 14:47 到每英里四美元的價格範圍 14:50 我們今天看到的 14:51 我們認為嗯，因為消費者 14:55 在西方市場，我們的分析顯示 14:57 他們估計自己的時間大概是 14:59 相當於 15:01 平均小時工資 15:03 行駛里程數，目的是 15:05 上班是為了通勤里程 15:08 大約相當於小時工資和 15:10 然後是非通勤里程或駕駛 15:13 您為個人使用而做的事情 15:14 這大約是每小時的一半 15:16 工資，所以這給了很多很多的需求 15:20 在西方市場的價格點 15:23 大約一美元 10 到 60 美分 15:26 然後在 15:28 我們預期的低收入國家 15:30 50 美分英里的巨大需求 15:33 mark um 那是因為如果你看 15:36 你知道deedy已經定價多少了 15:38 在像這樣的地方乘坐冰雹選擇 15:40 今天的中國已經很漂亮了 15:41 便宜約 50 美分，所以 15:44 最終每英里 25 美分的價格 15:46 點進來 15:47 我們認為這將是 15:49 額外客戶的長尾 15:52 現在將被帶入騎行 15:54 以前沒有的山市場 15:57 這就是你知道我們看到的 15:58 戲劇性的變化 16:00 在習慣和其他人中 16:02 為此進入市場 16:04 非常便宜的服務，所以最終 16:07 你知道所有這些價格點 16:08 考慮到我們的模型，呃，我們是 16:10 假設一旦特斯拉穿透 16:13 市場的某些部分，呃，如果他們 16:15 他們繼續增加自主里程 16:17 落入下一個呃價位 16:19 細分，這就是你最終的方式 16:20 可以考慮考慮 16:22 嗯 16:24 這 16:25 我們認為最終效果是什麼 16:27 有更多機會或更多收入 16:29 自動駕駛的機會比 16:31 我們以前認為有 16:33 所以當人們通過模型時，塔莎 16:36 和它一起玩 16:37 嗯，就像你要那樣做 16:39 你知道在 github 上下載這個，呃，但是 16:42 如果人們會這樣做，你知道並且 16:44 他們有問題或 16:46 呃，他們不明白的東西 16:47 人們最好的方法是什麼 16:50 伸出手與我們互動 16:51 因為你知道方舟的很大一部分 16:53 研究是開源的 16:56 和研究生態系統，那麼什麼是 16:58 人們獲得的最佳方式 17:01 參與和互動 17:03 好問題真的有兩種方法 17:05 對於我們的 github 書呆子，你可以 17:08 絕對在github上發帖 17:11 在那裡使用評論 17:14 提出您看到的任何問題 17:16 模型或問題部分 17:18 還向我們發推文你知道我是塔莎 17:22 弧 17:22 呃，山姆你的 17:25 句柄是 17:27 s 合唱弧 17:29 和我們的研究主管 b bewinton 17:31 弧 17:33 我們都可以回答問題 17:35 在模型上，我們很想听聽你的 17:37 回饋 17:39 arc 認為信息 17:41 呈現是準確的，並獲得 17:42 來自 arc 認為的來源 17:44 可靠但電弧不保證 17:46 任何的準確性或完整性 17:48 信息和此類信息可能是 17:50 如有更改，恕不另行通知 17:51 arc 歷史結果不是 17:53 未來結果的跡象 17:55 此播客中包含的聲明 17:57 可能是對未來預期的陳述 17:59 和其他前瞻性陳述 18:00 基於 arc 的當前視圖 18:02 和假設並涉及已知 18:03 未知的風險和不確定性 18:05 可能導致實際結果性能 18:07 或與有重大差異的事件 18:08 那些明示或暗示的 18:10 陳述 18:15 [音樂] 18:16 你
|
沒有留言:
張貼留言